The Webrary

Category: Guides

  • AI Ethics and Privacy: What Every User Should Know in 2026

    AI Ethics and Privacy: What Every User Should Know in 2026

    You paste confidential meeting notes into ChatGPT to get a summary. Your marketing team feeds customer data into an AI writing tool to personalize emails. Your developer asks Claude to debug production code containing API keys. Each of these actions has privacy implications that most users never consider.

    This guide covers what actually happens to your data when you use AI tools, where bias shows up in outputs, what the law says in 2026, and how to protect yourself and your organization.

    What Happens to Your Data: The Technical Reality

    When you send a prompt to an AI service, your data travels through several layers, each with different privacy characteristics.

    The Request Path

  • In transit: Your prompt is encrypted via TLS between your device and the provider’s servers. This is standard and all major providers do it. Man-in-the-middle attacks on properly configured HTTPS are not a practical concern.
  • At the server: Your prompt is processed by the model. During inference, your data exists in GPU memory temporarily. After the response is generated, the question is whether the provider retains your input and output.
  • In storage: This is where it matters. Different providers store your data for different durations and purposes. Some retain it for 30 days for abuse monitoring. Some retain it indefinitely for model training. Some delete it immediately after inference.
  • In training: The critical question — does your data get used to train future model versions? This varies by provider, plan tier, and configuration.

    • Provider-Specific Data Policies (as of Early 2026)

    OpenAI (ChatGPT, API)

    • Free/Plus users: By default, conversations are used to train models. You can opt out in Settings > Data Controls > “Improve the model for everyone,” but this is opt-out, not opt-in.
    • API users: Data is NOT used for training by default. Retained for 30 days for abuse monitoring, then deleted.
    • Enterprise/Team: Data is never used for training. SOC 2 compliant. Data retained per enterprise agreement.
    • Important caveat: Even with training opt-out enabled, OpenAI may review conversations flagged by automated systems for safety purposes. Human reviewers can see your content.

    Anthropic (Claude)

    • Free/Pro users: Conversations may be used for training unless you opt out. Anthropic’s data retention policy states conversations are kept for a limited period for safety and improvement.
    • API users: Data is not used for training by default. Retained for 30 days for trust and safety.
    • Enterprise: Data isolation, no training use, configurable retention periods.
    • Notable: Anthropic publishes detailed usage policies and has been more transparent than average about data handling practices.

    Google (Gemini)

    • Free Gemini users: Conversations are used to improve Google products, including model training. Data may be reviewed by human annotators. Retained for up to 3 years.
    • Workspace/Enterprise: Separate data processing agreements. Not used for training. Subject to enterprise data governance.
    • API (Vertex AI): Enterprise-grade data isolation. Not used for training.
    • Warning: Google’s consumer AI data policies are among the broadest. Free Gemini users should assume their conversations are not private.

    Microsoft (Copilot)

    • Consumer Copilot: Conversations may be used to improve Microsoft products. Data handling governed by Microsoft’s consumer privacy policy.
    • Copilot for Microsoft 365: Enterprise data protection. Queries processed within your Microsoft 365 tenant boundary. Not used for model training. Inherits your existing Microsoft 365 compliance certifications.

    The Rule of Thumb

    If you are using a free or consumer-tier AI product, assume your data is being stored and potentially used for training unless you have explicitly opted out. If privacy matters for your use case, use the API tier or enterprise plan, where data protections are contractually guaranteed rather than policy-based.

    Bias in AI Outputs: Where It Hides

    AI models reflect the biases present in their training data. This is not a theoretical concern — it has practical consequences in everyday use.

    Representation Bias

    Ask an image generation model to create “a CEO” and you will disproportionately get images of middle-aged white men. Ask a language model to write a story about “a nurse” and it will default to female pronouns more often than male. These biases mirror statistical distributions in training data (mostly internet text and images) rather than reflecting reality or ideals.

    Practical impact: If you use AI to generate marketing materials, job descriptions, or educational content without actively checking for representation bias, you may inadvertently reinforce stereotypes.

    Cultural and Geographic Bias

    Most major language models are trained predominantly on English-language, Western (especially American) internet content. This creates several blind spots:

    • Legal and regulatory advice defaults to US frameworks unless you specify otherwise.
    • Cultural norms in generated content reflect Western assumptions about business, social interactions, and communication styles.
    • Historical narratives tend toward Western perspectives on global events.
    • Language quality degrades for non-English outputs, with subtle errors in idiom, formality levels, and cultural context.

    Confirmation Bias in Research

    When you ask an AI to research a topic, it tends to generate balanced-sounding content that slightly favors the framing of your question. Ask “What are the benefits of remote work?” and you get a pro-remote-work summary. Ask “What are the problems with remote work?” and you get an anti-remote-work summary. Both sound authoritative. Neither tells you the model is giving you what you asked for rather than an objective analysis.

    Mitigation: Always ask the AI to present counterarguments to its own position. Request “steelman the opposing view” explicitly. Do not use AI research as a substitute for reading primary sources.

    Copyright and Intellectual Property

    The legal situation around AI-generated content is partially settled in 2026, but significant ambiguity remains.

    What Is Reasonably Clear

    AI-generated content is generally not copyrightable on its own. The US Copyright Office has maintained its position that works must have human authorship. Pure AI output — text or images generated with minimal human creative direction — does not qualify for copyright protection. This means your competitors can legally use your AI-generated marketing copy if they encounter it.

    Substantial human modification changes the equation. If you use AI to generate a first draft and then significantly rewrite, restructure, and add original analysis, the resulting work likely qualifies for copyright as a human-authored derivative work. The key factor is whether the human contribution is sufficient to constitute original authorship.

    Using copyrighted material in prompts is generally fine. Pasting a copyrighted article into an AI prompt for summarization or analysis is typically covered by fair use (in the US) — you are not reproducing the work publicly, you are processing it privately. However, if you then publish the AI’s summary, the analysis becomes more complex.

    What Remains Ambiguous

    Training data legality is still in active litigation. Multiple lawsuits (New York Times v. OpenAI, Getty Images v. Stability AI, and others) are challenging whether training AI models on copyrighted content constitutes fair use. Court decisions in late 2025 and early 2026 have been mixed, with no definitive Supreme Court ruling yet.

    AI-assisted invention patents remain a gray area. The USPTO has issued guidance that AI-assisted inventions can be patented if a human made a “significant contribution” to the invention, but the threshold for “significant” is not precisely defined.

    Liability for AI-generated misinformation is evolving. If your AI-powered tool generates defamatory content about a real person and you publish it, you are potentially liable — not the AI provider. Terms of service universally place responsibility for outputs on the user.

    Workplace AI Policies: What Your Company Needs

    If your organization uses AI tools and does not have a written policy, you are operating with uncontrolled risk. Here is what a functional AI usage policy should cover:

    Data Classification

    Define what data can and cannot be used with AI tools:

    • Unrestricted: Public information, general knowledge queries, non-sensitive creative tasks.
    • Internal only: Internal documents, meeting notes, project plans. Allowed only with enterprise-tier AI tools that guarantee no training use.
    • Confidential: Customer data, financial information, trade secrets, legal documents. Prohibited from external AI tools. Internal self-hosted models only, if at all.
    • Regulated: Data subject to HIPAA, PCI-DSS, GDPR, or similar regulations. Requires specific compliance verification before any AI processing.

    Disclosure Requirements

    Should employees disclose when content was AI-assisted? Best practice: yes, at least internally. This is not about shame — it is about quality control. Knowing which reports, analyses, and communications were AI-assisted helps reviewers calibrate their scrutiny. AI-generated financial projections need more verification than AI-generated meeting agendas.

    Approved Tools List

    Maintain a list of approved AI tools with their tier of use. Example:

    Tool Approved Use Data Level Allowed
    ChatGPT Enterprise General business use Internal
    Claude API Development, analysis Internal
    GitHub Copilot Business Code assistance Internal code only
    Jasper Business Marketing content Unrestricted
    Consumer ChatGPT/Claude Personal learning only Unrestricted

    Review and Accountability

    All AI-generated content published externally should be reviewed by a human who is accountable for its accuracy. “The AI wrote it” is not a defense for publishing incorrect information, defamatory statements, or regulatory violations.

    GDPR, the EU AI Act, and Global Regulations

    GDPR and AI (EU)

    GDPR applies to AI processing of personal data in straightforward ways:

    • Lawful basis: You need a legal basis (consent, legitimate interest, etc.) to process personal data through AI tools, just as you would with any other data processor.
    • Data processing agreements: If you use an AI API to process EU personal data, you need a DPA with the provider. Enterprise tiers from OpenAI, Anthropic, and Google offer these. Free tiers do not.
    • Right to explanation: If you make automated decisions that significantly affect individuals (hiring, credit, insurance), GDPR Article 22 gives those individuals the right to contest the decision and request human review.
    • Data minimization: Only send the minimum necessary personal data to AI tools. If you need to analyze customer feedback, anonymize names and identifying details before processing.

    EU AI Act (Enforcing 2026)

    The EU AI Act, with most provisions taking effect in 2026, classifies AI systems by risk level:

    • Unacceptable risk (banned): Social scoring by governments, real-time biometric surveillance in public spaces (with limited exceptions), manipulation of vulnerable groups.
    • High risk (heavily regulated): AI in hiring/recruitment, credit scoring, education assessment, law enforcement, critical infrastructure. Requires conformity assessments, human oversight, transparency, and logging.
    • Limited risk (transparency obligations): Chatbots must disclose they are AI. AI-generated content must be labeled when published in certain contexts (especially deepfakes).
    • Minimal risk (no specific requirements): Most consumer AI tools, creative assistants, productivity tools.

    Practical impact for most users: If you use AI tools for internal productivity (writing emails, summarizing documents, coding), you are in the minimal-risk category and face no new regulatory burden. If you use AI in hiring, customer-facing decisions, or content generation that could be mistaken for human-created journalism, you need to check your compliance obligations.

    United States

    The US has no comprehensive federal AI regulation as of early 2026. Regulation is fragmented across:

    • Executive orders establishing AI safety guidelines for federal agencies
    • State laws (Colorado’s AI Act, California’s proposed AI transparency requirements)
    • Sector-specific guidance from FTC (deceptive practices), FDA (medical AI), SEC (financial AI)
    • FTC enforcement against companies making misleading AI claims

    The practical effect is that US-based users have fewer hard legal requirements but more legal uncertainty. Follow FTC guidelines on transparency and avoid using AI in ways that could be considered deceptive or unfair.

    Practical Tips for Safe AI Usage

    These are not theoretical suggestions — they are habits that prevent real problems.

    1. Never paste credentials, API keys, passwords, or tokens into AI prompts. This seems obvious, but developers do it constantly when asking AI to debug configuration files. Strip sensitive values before pasting. Use placeholder text like YOUR_API_KEY_HERE.

    2. Anonymize personal data before processing. If you need AI to analyze customer support tickets, replace names, email addresses, phone numbers, and account numbers with pseudonyms first. Many organizations automate this with regex-based scrubbing scripts.

    3. Verify every factual claim in AI output. AI models hallucinate — they generate confident, specific, false information. Statistics, dates, quotes, citations, and technical specifications are the most common hallucination categories. Never publish AI-generated factual claims without independent verification.

    4. Use separate accounts for personal and professional AI use. Your personal ChatGPT conversation about vacation planning should not share a context with your professional conversations about quarterly revenue.

    5. Check the training data opt-out settings every time you update an app or change your subscription tier. Providers occasionally reset preferences during updates. Verify your settings monthly.

    6. Download and review your data periodically. OpenAI, Google, and Anthropic all offer data export features. Review what they have stored about you and delete what you do not want retained.

    7. Do not use AI for high-stakes decisions without human oversight. Hiring decisions, medical interpretations, legal advice, financial recommendations — these all require human judgment and accountability. AI can assist but should not decide.

    AI Tool Privacy Evaluation Checklist

    Before adopting any AI tool for professional use, evaluate it against these criteria:

    Data Handling

    • [ ] Does the provider clearly state whether your data is used for training?
    • [ ] Can you opt out of training data use?
    • [ ] What is the data retention period?
    • [ ] Is data encrypted at rest and in transit?
    • [ ] Where are the servers located (relevant for data residency requirements)?

    Compliance

    • [ ] Does the provider offer a Data Processing Agreement?
    • [ ] Is the service SOC 2 Type II certified?
    • [ ] Does it comply with GDPR (if processing EU data)?
    • [ ] Does it meet your industry-specific requirements (HIPAA, PCI-DSS, etc.)?

    Access Control

    • [ ] Can you control which team members have access?
    • [ ] Are conversation logs accessible to administrators?
    • [ ] Can you set data classification restrictions per user or team?

    Transparency

    • [ ] Does the provider publish a transparency report?
    • [ ] Are there clear terms about when human reviewers can access your data?
    • [ ] Does the provider notify you of policy changes?

    Incident Response

    • [ ] Does the provider have a documented data breach notification process?
    • [ ] What is the notification timeline (GDPR requires 72 hours)?
    • [ ] Is there a dedicated security contact?

    If an AI tool cannot satisfy the data handling and compliance sections of this checklist, do not use it for any data beyond publicly available information.

    The Bottom Line

    AI ethics and privacy are not abstract philosophical topics — they are practical risk management. Every time you interact with an AI tool, you are making decisions about data exposure, bias propagation, intellectual property, and regulatory compliance. The organizations and individuals who thrive in the AI era will be those who use these tools aggressively while managing their risks deliberately.

    Start with your data classification. Audit your current AI tool usage against the checklist above. Write or update your organization’s AI policy. And build the habit of pausing for two seconds before pasting anything into an AI prompt to ask: “Would I be comfortable if this appeared in a training dataset?”

    That two-second habit is worth more than any privacy policy.

  • AI Tools for Small Business: A Practical Guide to Getting Started

    AI Tools for Small Business: A Practical Guide to Getting Started

    AI is not just for tech giants anymore. A bakery owner can automate customer emails. A landscaping company can generate quotes in seconds instead of hours. A boutique retailer can create social media content without hiring a designer. The tools exist today, they are affordable, and most of them do not require any technical knowledge to set up.

    But the hype makes it hard to separate genuinely useful tools from expensive toys. This guide focuses exclusively on AI tools that deliver measurable time savings or revenue improvements for small businesses — the kind with 1 to 50 employees, limited budgets, and no dedicated IT staff.

    Customer Service: Respond Faster Without Hiring

    Customer service is where most small businesses feel AI’s impact first. Responding to the same questions repeatedly — business hours, pricing, return policies, appointment availability — is exactly the kind of repetitive work that AI handles well.

    AI Chatbots for Your Website

    What they do: Answer customer questions automatically, 24/7, using information you provide about your business. Modern chatbots understand natural language — customers can ask questions in their own words, not just click pre-written options.

    Best tools:

    • Tidio ($29/month for the AI plan) — Connects to your website in minutes. You feed it your FAQ, pricing page, and policies. It handles roughly 70% of incoming questions without human intervention. When it cannot answer, it collects the customer’s information and alerts you. Works with Shopify, WordPress, and most website builders.
    • Intercom Fin ($0.99 per resolved conversation) — More sophisticated but pricier. Fin reads your entire help center and resolves conversations autonomously. The per-resolution pricing means you only pay when it actually helps someone. Good for businesses with 50+ customer interactions per day.
    • ChatGPT with a custom GPT (ChatGPT Plus, $20/month) — The budget option. Create a custom GPT trained on your business information and share the link with customers. It lacks the polish of dedicated chatbot platforms (no website widget, no handoff to humans), but costs a fraction of the price.

    ROI calculation: If you spend 2 hours per day answering routine customer questions, and a chatbot handles 70% of them, you save roughly 1.4 hours daily. At a $25/hour labor value, that is $35/day or approximately $1,050/month — far exceeding the cost of any chatbot tool.

    AI Email Management

    What it does: Drafts replies to customer emails, categorizes incoming messages by urgency, and flags messages that need personal attention.

    Best tools:

    • Superhuman AI ($30/month) — Drafts email replies in your writing style after learning from your sent messages. The time savings compound: instead of writing 30 emails from scratch, you review and send 30 AI-drafted emails. Most need only minor edits.
    • Gmail’s built-in AI (included with Google Workspace, $7+/month) — Google’s “Help me write” feature drafts replies and composes new emails. Less sophisticated than Superhuman but already included if you use Google Workspace.

    Time saved: Most users report cutting email time by 40-60%. For a business owner spending 1 hour per day on email, that is 25-35 minutes reclaimed daily.

    Marketing: Create Content Without a Creative Team

    Marketing is the area where AI tools have matured the most. Content that used to require a copywriter, designer, and social media manager can now be produced by a single person using AI assistance.

    Content Writing

    What it does: Generates blog posts, product descriptions, ad copy, newsletters, and social media captions. The best tools produce drafts that need editing, not rewriting.

    Best tools:

    • Claude ($20/month for Pro) — Excels at longer-form content like blog posts, newsletters, and detailed product descriptions. Produces notably natural-sounding copy that requires less editing than competitors. Strong at matching your brand voice when given examples.
    • Jasper ($49/month) — Built specifically for marketing teams. Includes templates for ads, emails, landing pages, and social posts. The brand voice feature learns your style and maintains consistency across all content. More expensive but saves time with its structured templates.
    • ChatGPT Plus ($20/month) — The most versatile option. Handles everything from social captions to long-form articles. Lacks marketing-specific templates but makes up for it with flexibility.

    Practical workflow: Do not ask AI to write your entire blog post from scratch and publish it as-is. Instead: (1) brainstorm topics with AI, (2) create an outline together, (3) draft each section with AI assistance, (4) edit heavily for your voice and expertise, (5) add your own examples and experiences. The result is authentic content produced in a third of the time.

    ROI calculation: A professional copywriter charges $50-150/hour. If you produce 4 blog posts per month (8 hours of writing time), AI tools reduce that to 3 hours of writing and editing. At $75/hour copywriter rates, you save $375/month while maintaining quality through your own editorial oversight.

    Social Media Content

    What it does: Generates post captions, suggests content calendars, creates image variations, and repurposes existing content across platforms.

    Best tools:

    • Canva Magic Studio (included with Canva Pro, $13/month) — Generates social media graphics with AI, removes backgrounds, resizes designs for different platforms, and writes captions. For small businesses already using Canva, this is the highest-value upgrade.
    • Buffer AI Assistant (included with Buffer paid plans, $6+/month) — Generates post ideas and captions directly in your scheduling workflow. Suggests optimal posting times. Less powerful than dedicated AI tools but eliminates the friction of switching between apps.
    • Opus Clip ($19/month) — Takes long-form video (a webinar, interview, or product demo) and automatically clips it into short-form content for TikTok, Instagram Reels, and YouTube Shorts. Identifies the most engaging moments and adds captions. If you produce any video content, this tool pays for itself immediately.

    Time saved: Creating a week’s worth of social media content typically drops from 4-6 hours to 1-2 hours. The AI handles the first draft of every caption and suggests visual concepts; you refine and approve.

    Email Marketing

    What it does: Writes email sequences, subject lines, and newsletter content. Some tools also optimize send times and segment your audience.

    Best tools:

    • Mailchimp AI (included with Standard plan, $20/month) — Generates email content, suggests subject lines, and optimizes send times based on your audience’s behavior. The subject line generator alone improves open rates measurably — it A/B tests AI-generated variations automatically.
    • Klaviyo AI (free up to 250 contacts) — Specifically designed for e-commerce. Generates product recommendation emails, abandoned cart sequences, and win-back campaigns. The AI segments your audience based on purchasing behavior and personalizes content for each segment.

    Operations: Automate the Tedious Work

    Operational tasks — scheduling, inventory tracking, data entry — consume hours that small business owners could spend on growth. AI tools in this category are less flashy than marketing tools but often deliver the highest ROI.

    Scheduling and Appointments

    What it does: Handles appointment booking, sends reminders, manages cancellations, and optimizes your calendar.

    Best tools:

    • Reclaim.ai ($10/month) — AI-powered calendar management that automatically finds time for tasks, meetings, and breaks. It learns your preferences (no meetings before 10 AM, focused work in the morning) and defends your time. Particularly valuable for service businesses juggling client appointments with operational work.
    • Calendly with AI ($12/month) — The booking tool you probably already know, now with AI features that suggest optimal meeting lengths, detect scheduling conflicts, and automate follow-up messages.

    Document Processing

    What it does: Extracts information from invoices, receipts, contracts, and forms. Eliminates manual data entry.

    Best tools:

    • Docsumo (from $50/month) — Extracts data from invoices, purchase orders, and bank statements with 98%+ accuracy. Connects to QuickBooks, Xero, and other accounting software. If you process more than 50 documents per month, the time savings justify the cost.
    • Adobe Acrobat AI Assistant (included with Acrobat Pro, $23/month) — Summarizes long documents, answers questions about contract terms, and extracts key data points. Useful for businesses that deal with contracts, legal documents, or lengthy vendor agreements.

    Inventory and Supply Chain

    What it does: Predicts demand, suggests reorder points, and identifies slow-moving stock.

    Best tools:

    • inFlow ($110/month for the AI features) — Inventory management with demand forecasting. Analyzes your sales history and predicts what you will need to reorder and when. Reduces both stockouts and overstock situations. The ROI is significant for product-based businesses: carrying excess inventory costs 20-30% of the inventory value per year.
    • Shopify’s built-in AI (included with Shopify plans) — If you sell through Shopify, the built-in inventory predictions and demand forecasting handle the basics without an additional tool.

    Finance: Smarter Bookkeeping and Forecasting

    Bookkeeping Automation

    What it does: Categorizes transactions, reconciles accounts, flags anomalies, and reduces the time your bookkeeper or accountant spends on routine tasks.

    Best tools:

    • QuickBooks AI (included with QuickBooks Online, $35+/month) — Auto-categorizes bank transactions with improving accuracy over time. Flags unusual transactions for review. Generates cash flow forecasts based on your historical patterns. If you already use QuickBooks, these features activate automatically.
    • Vic.ai (custom pricing, typically $200+/month) — Enterprise-grade accounts payable automation. Processes invoices, matches them to purchase orders, and routes them for approval. Overkill for most small businesses, but transformative for companies processing 100+ invoices monthly.

    Financial Forecasting

    What it does: Projects revenue, expenses, and cash flow based on your historical data and market trends.

    Best tools:

    • Fathom ($49/month, connects to QuickBooks/Xero) — Generates visual financial reports and forecasts. The AI identifies trends in your financials and alerts you to potential problems (declining margins, seasonal cash flow gaps) before they become critical.
    • Float ($59/month) — Cash flow forecasting that connects to your accounting software. Shows you exactly when cash will be tight and suggests actions (delay a purchase, accelerate an invoice) to stay healthy. For businesses that have experienced cash flow surprises, this tool provides genuine peace of mind.

    Implementation Roadmap: Start Here

    Do not try to adopt everything at once. Follow this phased approach:

    Month 1: Quick Wins (Budget: $20-50/month)

    Start with tools that save time immediately with minimal setup:

  • Sign up for ChatGPT Plus or Claude Pro ($20/month). Use it for email drafting, content writing, and brainstorming. Spend the first week learning to write effective prompts for your specific needs.
  • Set up a basic chatbot on your website using Tidio’s free tier. Configure it with your top 10 FAQs. Monitor the conversations it handles and refine its responses weekly.
  • Measure your baseline. Track how much time you spend on the tasks you are automating. You need this data to calculate ROI later.

    • Month 2: Marketing Acceleration (Budget: $50-100/month)

    Once you are comfortable with the basics:

  • Upgrade Canva to Pro ($13/month) and start using Magic Studio for social media graphics.
  • Create a content calendar with AI assistance. Generate a month’s worth of blog post outlines and social media captions in one focused session.
  • Set up an email sequence in your email marketing platform using AI-generated content. Start with a welcome sequence for new subscribers — it runs on autopilot once created.

    • Month 3: Operations Optimization (Budget: $100-200/month)

    Now tackle the operational bottlenecks:

  • Implement scheduling automation with Reclaim.ai or Calendly if you handle appointments.
  • Connect document processing if you handle significant paperwork. Start with invoice processing — it has the clearest ROI.
  • Review your financial tools and add forecasting if your current accounting software lacks it.

    • Ongoing: Measure and Adjust

    After three months, calculate your actual ROI:

    Most small businesses find a 3-5x return on their AI tool spending within the first quarter. The businesses that see the highest ROI are those that consistently use the tools daily rather than setting them up and forgetting about them.

    Common Mistakes to Avoid

    Buying tools before identifying the problem. Start with your biggest time sinks, then find tools that address them. Do not subscribe to an AI tool because it looks impressive — subscribe because it solves a specific problem you have.

    Expecting perfection from day one. AI tools improve as you use them. Chatbots get better as you refine their knowledge base. Writing tools produce better content as you learn to prompt them effectively. Give each tool at least 2-3 weeks of consistent use before judging it.

    Skipping the human review. AI-generated content, emails, and customer responses should always be reviewed before they go out. The tool produces the first draft; you provide the quality control, personal touch, and brand voice. Fully automated customer-facing content without human review is how businesses damage their reputation.

    Ignoring your team. If you have employees, involve them in choosing and implementing AI tools. The person who answers customer emails daily will have better insight into what a chatbot should handle than someone who reads about it in a blog post. Adoption succeeds when the people using the tools have a say in selecting them.

    The Bottom Line

    AI tools for small business are not about replacing people — they are about giving your existing team (even if that team is just you) the ability to accomplish more in less time. Start with one tool that addresses your most painful time sink. Learn it well. Measure the results. Then expand to the next area. Within three months, you will have a clear picture of which tools earn their cost and which do not. The investment is small; the time savings are substantial; and the competitive advantage of moving early is real.

  • AI Agents Explained: How Autonomous AI Systems Actually Work

    AI Agents Explained: How Autonomous AI Systems Actually Work

    The term “AI agent” has become one of the most overused buzzwords in tech. Every startup claims to have one, every framework promises to help you build one, and every demo looks impressive until you try to use it on real work. This guide strips away the marketing and explains what AI agents actually are, how they work architecturally, what they can and cannot do today, and how to build a simple one yourself.

    What Is an AI Agent? A Clear Definition

    An AI agent is a software system that uses a language model to autonomously decide what actions to take in order to accomplish a goal. The key word is autonomously — unlike a chatbot that responds to a single prompt and stops, an agent operates in a loop: it observes its environment, reasons about what to do next, takes an action, observes the result, and repeats until the goal is achieved or it determines it cannot proceed.

    The distinction matters. When you ask ChatGPT to “write a blog post,” that is a single-turn interaction — not an agent. When you ask a system to “research competitor pricing, create a comparison spreadsheet, and draft a summary email,” and it breaks that into sub-tasks, executes each one using different tools, handles errors along the way, and delivers the final result — that is an agent.

    Three properties define a true agent:

  • Autonomy: It decides its own next steps rather than following a fixed script.
  • Tool use: It can interact with external systems — APIs, databases, file systems, browsers, code interpreters.
  • Persistence: It maintains state across multiple steps, remembering what it has done and what it still needs to do.

    • The Architecture: Perception-Reasoning-Action Loop

    Every AI agent, regardless of framework or complexity, follows the same fundamental loop:

    1. Perception (Observe)

    The agent receives input about its current state. This can include:

    2. Reasoning (Think)

    The language model processes all available context and decides what to do next. This is where the “intelligence” lives. The model evaluates:

    Modern agents often use structured reasoning techniques. Chain-of-thought prompting forces the model to articulate its reasoning before deciding on an action, which significantly reduces errors. Some frameworks implement explicit “scratchpad” areas where the model writes out its thinking.

    3. Action (Do)

    The agent executes the chosen action through a tool. Common tool categories include:

    4. Observation (Check)

    The agent receives the result of its action and feeds it back into the perception step. The loop continues until one of three conditions is met:

    Types of AI Agents

    Not all agents are built the same. The architecture varies based on the complexity of the task and the level of autonomy required.

    Reactive Agents

    The simplest type. A reactive agent responds directly to the current input without maintaining an internal model of the world. Think of a customer support bot that routes queries to the right department based on keywords — it makes decisions but does not plan ahead or remember previous interactions in a meaningful way.

    Strengths: Fast, predictable, easy to debug.
    Weaknesses: Cannot handle multi-step tasks, no learning, no planning.

    Deliberative Agents (Plan-and-Execute)

    These agents create an explicit plan before taking any action. They break the goal into sub-tasks, determine the order of execution, and then work through the plan step by step. If a step fails, they can re-plan.

    This is the architecture used by most production agent systems today. The planning step adds latency but dramatically improves reliability on complex tasks.

    Strengths: Handles complex, multi-step tasks. Can recover from failures.
    Weaknesses: Planning adds latency. Plans can be wrong, leading to wasted effort before re-planning.

    Multi-Agent Systems

    Instead of one agent handling everything, multi-agent systems assign different agents to different roles. A “manager” agent might decompose a task and delegate sub-tasks to specialized agents — one for research, one for writing, one for code review.

    This architecture mirrors how human teams work and can outperform single agents on complex projects. However, coordination overhead is real: agents need to communicate effectively, avoid duplicate work, and resolve conflicts when their outputs contradict each other.

    Strengths: Parallel execution, specialized expertise per agent, better for large tasks.
    Weaknesses: Complex to orchestrate, communication overhead, harder to debug.

    Real-World AI Agents in 2026

    AutoGPT and Open-Source Pioneers

    AutoGPT (launched 2023) was the first widely-known autonomous agent. It demonstrated the concept of an AI that could browse the web, write files, and execute code to accomplish goals. The initial versions were unreliable — they would get stuck in loops, waste API credits on circular reasoning, and frequently fail on tasks that seemed simple.

    By 2026, the descendants of AutoGPT (including AgentGPT, BabyAGI, and various forks) have improved significantly. Better models, structured output formats, and more robust tool implementations have made open-source agents genuinely useful for certain tasks like research synthesis and data analysis.

    Devin (Cognition)

    Devin positioned itself as an “AI software engineer” capable of handling entire development tasks: reading codebases, planning implementations, writing code, running tests, and debugging failures. The reality is more nuanced — Devin works well on well-defined, isolated tasks (fix this bug, add this feature to this file) but struggles with ambiguous requirements, large-scale architectural decisions, and tasks that require deep understanding of business context.

    What Devin got right was the tool integration. It operates in a full development environment with a shell, browser, code editor, and terminal, giving it the same tools a human developer uses.

    Claude Computer Use (Anthropic)

    Anthropic’s computer use capability lets Claude interact with a computer through screenshots and mouse/keyboard actions — essentially using a computer the way a human does. This is a fundamentally different approach from API-based tool use. Instead of calling a structured function, the agent looks at the screen, decides where to click, types text, and observes the result.

    The advantage is universality: any application with a GUI becomes a “tool” without building custom integrations. The disadvantage is speed and reliability — clicking through UI elements is slower than API calls and more prone to errors from layout changes or unexpected popups.

    OpenAI Operator

    OpenAI’s Operator focuses on web-based tasks: booking reservations, filling out forms, navigating websites, and completing multi-step online workflows. It combines browsing capabilities with structured reasoning to handle tasks that previously required browser automation scripts (like Selenium or Playwright) but with the flexibility to handle unexpected page layouts.

    Operator works best for repetitive web tasks with clear success criteria. It struggles with tasks requiring judgment calls, ambiguous instructions, or websites with aggressive bot detection.

    Tool Use and Function Calling: The Engine Room

    The practical power of an agent comes from its tools. Here is how tool use works under the hood.

    When you define a tool for an agent, you provide:

  • A name: What the tool is called (e.g., search_web, read_file, send_email)
  • A description: What the tool does, so the model knows when to use it
  • A parameter schema: What inputs the tool accepts, in JSON Schema format
  • An implementation: The actual code that runs when the tool is called

    • The language model does not execute the tool directly. It outputs a structured request (typically JSON) specifying which tool to call and with what parameters. The agent framework intercepts this, executes the tool, and feeds the result back to the model.

    # Example tool definition for an agent
tools = [
    {
        "name": "search_web",
        "description": "Search the web for current information. Use when you need facts, data, or recent events.",
        "parameters": {
            "type": "object",
            "properties": {
                "query": {
                    "type": "string",
                    "description": "The search query"
                }
            },
            "required": ["query"]
        }
    },
    {
        "name": "read_url",
        "description": "Read the full text content of a web page.",
        "parameters": {
            "type": "object",
            "properties": {
                "url": {
                    "type": "string",
                    "description": "The URL to read"
                }
            },
            "required": ["url"]
        }
    }
]

    The quality of your tool descriptions directly impacts agent performance. Vague descriptions lead to tools being used inappropriately. Overly restrictive descriptions cause the agent to avoid useful tools. Write descriptions as if you are explaining the tool to a competent colleague who has never seen it before.

    Memory Systems: Short-Term and Long-Term

    Agents need memory to function across multiple steps and sessions.

    Short-term memory is the conversation context — everything the agent has seen and done in the current session. This is limited by the model’s context window. For a complex task with many tool calls, you can exhaust context quickly. Strategies to manage this include summarizing previous steps, dropping tool outputs after they have been processed, and compressing conversation history.

    Long-term memory persists across sessions. Implementations include:

    Memory is still one of the weakest aspects of current agent systems. Most agents in 2026 have functional short-term memory and rudimentary long-term memory at best.

    Building a Simple Agent: Working Code

    Here is a complete, minimal agent using Python and the OpenAI API that can search the web and answer questions:

    import json
import openai
import requests

client = openai.OpenAI()


    Tool implementations
    
def search_web(query: str) -> str:
    """Search using a search API and return results."""
    # Using a hypothetical search API; replace with your preferred provider
    response = requests.get(
        "https://api.search.example/v1/search",
        params={"q": query, "num": 5},
        headers={"Authorization": "Bearer YOUR_API_KEY"}
    )
    results = response.json().get("results", [])
    return "n".join(
        f"- {r['title']}: {r['snippet']} ({r['url']})"
        for r in results
    )

def calculate(expression: str) -> str:
    """Safely evaluate a mathematical expression."""
    try:
        # Only allow safe math operations
        allowed = set("0123456789+-*/.() ")
        if all(c in allowed for c in expression):
            return str(eval(expression))
        return "Error: Invalid expression"
    except Exception as e:
        return f"Error: {e}"

TOOLS = {
    "search_web": search_web,
    "calculate": calculate,
}

TOOL_SCHEMAS = [
    {
        "type": "function",
        "function": {
            "name": "search_web",
            "description": "Search the web for current information.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {"type": "string", "description": "Search query"}
                },
                "required": ["query"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "calculate",
            "description": "Calculate a mathematical expression.",
            "parameters": {
                "type": "object",
                "properties": {
                    "expression": {"type": "string", "description": "Math expression"}
                },
                "required": ["expression"]
            }
        }
    }
]

def run_agent(goal: str, max_steps: int = 10):
    messages = [
        {"role": "system", "content": (
            "You are a helpful research agent. Use the available tools to "
            "answer the user's question accurately. Think step by step. "
            "When you have enough information, provide a final answer."
        )},
        {"role": "user", "content": goal}
    ]
    
    for step in range(max_steps):
        response = client.chat.completions.create(
            model="gpt-4o",
            messages=messages,
            tools=TOOL_SCHEMAS,
            tool_choice="auto"
        )
        
        message = response.choices[0].message
        messages.append(message)
        
        # If no tool calls, the agent is done
        if not message.tool_calls:
            print(f"nFinal answer:n{message.content}")
            return message.content
        
        # Execute each tool call
        for tool_call in message.tool_calls:
            func_name = tool_call.function.name
            args = json.loads(tool_call.function.arguments)
            
            print(f"Step {step + 1}: Calling {func_name}({args})")
            
            result = TOOLSfunc_name
            
            messages.append({
                "role": "tool",
                "tool_call_id": tool_call.id,
                "content": result
            })
    
    return "Max steps reached without completing the task."


    Usage
    
answer = run_agent("What is the current population of Tokyo and how does it compare to New York City?")

    This is roughly 80 lines of code and implements a functional agent with tool use, multi-step reasoning, and a safety limit. Production agents add error handling, retry logic, logging, cost tracking, and more sophisticated memory management — but the core loop is identical.

    Current Limitations: What Agents Cannot Do Yet

    Reliability: Even the best agents fail 20-40% of the time on complex tasks. They get stuck in loops, misinterpret tool outputs, make incorrect assumptions, and occasionally hallucinate tool calls that do not exist. This makes agents unsuitable for fully unsupervised critical tasks.

    Cost: A single agent run can consume dozens of API calls. A complex research task might cost $1-5 in API credits — acceptable for high-value tasks but prohibitive at scale for low-value automation.

    Speed: Agent loops are inherently serial. Each step requires a full LLM inference pass plus tool execution time. A 10-step task might take 30-60 seconds, compared to sub-second responses for single-turn interactions.

    Context limits: Long-running agents accumulate context quickly. Tool outputs, intermediate results, and conversation history fill the context window, eventually forcing the agent to operate with incomplete information.

    Security: Giving an agent access to tools means giving it access to your systems. A misconfigured agent with file write access and internet connectivity could exfiltrate data, modify files destructively, or run expensive operations. Always sandbox agent tools and implement permission boundaries.

    The Future: What Is Coming Next

    The trajectory is clear even if the timeline is uncertain. Expect these developments over the next 12-18 months:

    Longer context and better memory will allow agents to work on tasks spanning hours or days rather than minutes. Models with 1M+ token context windows are already emerging, and structured memory systems are improving rapidly.

    Better tool ecosystems will reduce the integration work required to connect agents to real systems. Standardized tool protocols (like Anthropic’s Model Context Protocol) will make tools interoperable across agent frameworks.

    Multi-modal agents that can see, hear, and interact with GUIs will expand the range of tasks agents can handle without custom API integrations.

    Agent-to-agent communication standards will enable complex workflows where specialized agents collaborate on tasks too large for any single agent.

    The agents of 2026 are roughly where web applications were in 2005 — clearly useful, sometimes frustrating, and improving fast enough that today’s limitations will look quaint in two years. Start learning to build and use them now, but keep your expectations calibrated to current reality rather than future potential.

  • The Ultimate Guide to AI Image Generators: From DALL-E to Stable Diffusion

    The Ultimate Guide to AI Image Generators: From DALL-E to Stable Diffusion

    AI image generation has moved from a novelty to a practical creative tool. Designers use it for concept art, marketers generate social media visuals, developers create placeholder assets, and entire illustration workflows now start with an AI-generated base. But the market is fragmented — each tool has different strengths, pricing models, and licensing terms.

    This guide covers how these tools actually work, compares the top options head-to-head, teaches you to write prompts that produce consistent results, and addresses the commercial licensing question that trips up most newcomers.

    How AI Image Generation Works (Without the Math)

    All modern image generators are based on a technique called diffusion. Understanding the basics will make you better at prompting.

    Imagine starting with a photograph and gradually adding random noise until the image becomes pure static — like TV snow. A diffusion model learns to reverse this process. Given pure noise, it can progressively remove the noise to reveal a coherent image. The text prompt guides this denoising process, steering the output toward images that match your description.

    This is why diffusion models are surprisingly good at composition and style but struggle with certain things:

    • They excel at: textures, lighting, atmosphere, artistic styles, and spatial composition. These are properties the model learns deeply from its training data.
    • They struggle with: exact counts of objects, readable text in images, precise spatial relationships (“the red ball is exactly between the two blue cups”), and consistent human hands. These require precise symbolic reasoning that the denoising process handles imperfectly.

    Understanding these strengths and limitations directly improves your prompting strategy. Lean into what diffusion does well; work around what it does not.

    Comparing the Top Tools

    DALL-E 3 (OpenAI)

    Access: ChatGPT Plus ($20/month), API
    Resolution: Up to 1024×1792
    Speed: 10-20 seconds per image

    DALL-E 3 is the most accessible option because it is built into ChatGPT. You describe what you want in natural language, and ChatGPT actually rewrites your prompt behind the scenes to be more detailed and specific before sending it to the image model. This “prompt rewriting” is both its biggest strength and its most frustrating limitation.

    Strengths: DALL-E 3 handles complex prompts with multiple elements better than most competitors. “A golden retriever wearing a tiny chef hat, cooking pasta in a rustic Italian kitchen, warm afternoon light through the window” produces coherent, well-composed results consistently. Text rendering in images is also significantly better than other tools — it can put readable words on signs, book covers, and labels.

    Limitations: You have limited control over the exact aesthetic. The prompt rewriting system sometimes overrides your intent, adding details you did not ask for or interpreting your description differently than expected. There is no negative prompting (telling it what to exclude), and no way to control specific generation parameters like sampling steps or guidance scale.

    Best for: Quick concept generation, images that need readable text, non-technical users who want results without learning prompting syntax.

    Midjourney

    Access: Subscription ($10-60/month), Discord or web interface
    Resolution: Up to 2048×2048 (with upscaling)
    Speed: 30-60 seconds per image

    Midjourney produces the most aesthetically polished images of any generator. Its default style has a distinctive quality — rich colors, dramatic lighting, and a painterly feel that makes outputs look “finished” without extensive prompting.

    Strengths: The aesthetic quality ceiling is the highest in the industry. Midjourney excels at cinematic compositions, architectural visualization, character design, and anything where visual beauty matters more than photographic accuracy. Version 6.1 brought major improvements to photorealism, and the results can be genuinely difficult to distinguish from professional photography in many categories.

    The --style and --stylize parameters give you a slider between “follow my prompt exactly” and “make it beautiful.” The --chaos parameter introduces variation between outputs, useful when exploring ideas. Multi-prompt weighting with :: syntax lets you control the relative importance of different elements.

    Prompt tip: Midjourney responds exceptionally well to photography terminology. “85mm lens, f/1.4, golden hour, bokeh background” produces dramatically different results than the same subject without these terms. Mentioning specific artists, art movements, or visual styles also has a strong effect.

    Limitations: Until recently, Midjourney was Discord-only, which made it awkward for professional workflows. The web interface improves this but is still maturing. There is no API for programmatic access, which rules it out for automated pipelines. Prompt iteration is slower than API-based tools because you wait for the Discord bot or web UI.

    Best for: Marketing visuals, concept art, any use case where aesthetic quality is the primary concern.

    Stable Diffusion (Stability AI)

    Access: Free (open source), or Stability AI API
    Resolution: Configurable, typically 512×512 to 2048×2048
    Speed: 5-30 seconds depending on hardware

    Stable Diffusion is the open-source option, and that changes everything about how you use it. You can run it on your own GPU, fine-tune it on custom datasets, and integrate it into any pipeline without per-image costs.

    Strengths: Complete control. You can adjust every parameter: sampling method, guidance scale, steps, seed, and scheduler. ControlNet extensions let you guide generation with edge maps, depth maps, pose skeletons, and more — producing results that match a specific composition precisely. LoRA fine-tuning lets you train the model on a specific style, character, or product with as few as 20 reference images.

    SDXL and SD3 brought quality on par with commercial options for most use cases. The community has produced thousands of fine-tuned models for specific styles — anime, photorealism, architectural rendering, pixel art — each outperforming the base model in its niche.

    Limitations: The learning curve is steep. Getting started requires either a capable GPU (8GB+ VRAM recommended, 12GB+ preferred) or using a cloud GPU service. The tooling ecosystem (ComfyUI, Automatic1111, Forge) is powerful but intimidating for newcomers. Without fine-tuning or careful prompting, default quality lags behind Midjourney’s polished output.

    Best for: Developers building image generation into products, teams needing high-volume generation without per-image costs, anyone who needs fine-tuned models or precise composition control.

    Flux (Black Forest Labs)

    Access: Open source (Flux.1 Schnell/Dev), API (Flux Pro)
    Resolution: Up to 2048×2048
    Speed: 2-8 seconds (Schnell), 10-20 seconds (Pro)

    Flux emerged as a serious contender by offering Midjourney-tier quality in an open-source package. Built by former Stability AI researchers, it uses a more efficient architecture that produces high-quality images with fewer steps, meaning faster generation.

    Strengths: Flux.1 Schnell (the fast, open variant) generates usable images in 1-4 steps — dramatically faster than Stable Diffusion’s typical 20-30 steps. This makes it practical for real-time or near-real-time applications. Text rendering is surprisingly good for an open model. Flux Pro, the commercial API, produces results that consistently rival Midjourney in blind comparisons.

    Limitations: The ecosystem is younger than Stable Diffusion’s. Fewer LoRAs, fewer community models, and less mature tooling. ControlNet equivalents exist but are less battle-tested. The open-source variants (Schnell and Dev) have different licenses — Schnell is Apache 2.0 (truly open), while Dev is non-commercial.

    Best for: Applications needing fast generation, developers wanting open-source quality close to commercial tools, real-time creative tools.

    Ideogram

    Access: Free tier + subscriptions ($8-48/month)
    Resolution: Up to 1024×1024
    Speed: 15-30 seconds

    Ideogram carved out a niche with one specific capability: it renders text in images more accurately than any other tool. If you need a poster, logo mockup, or social media graphic with readable typography, Ideogram is the strongest choice.

    Strengths: Text rendering is Ideogram’s standout feature. “A vintage coffee shop sign that says ‘The Daily Grind’” produces an image where the text is actually legible and stylistically appropriate. Other tools either garble the text or render it as illegible shapes. The general image quality is competitive, though not best-in-class for non-text imagery.

    Limitations: Outside of text-heavy images, Ideogram does not match Midjourney’s aesthetic quality or Stable Diffusion’s flexibility. The API is limited, and the ecosystem is small.

    Best for: Marketing materials with text, logo concepts, signage mockups, social media graphics, any image where readable text is essential.

    Prompt Crafting: Techniques That Actually Work

    Good prompting is the difference between “that is sort of what I wanted” and “that is exactly right.” Here are techniques that produce consistent results across all tools.

    Structure Your Prompts in Layers

    Think of your prompt as having four layers:

  • Subject: What is in the image. “A calico cat sitting on a windowsill.”
  • Environment: Where the subject exists. “In a sun-drenched Parisian apartment, white curtains billowing.”
  • Style: How it should look. “Watercolor illustration, soft edges, muted warm palette.”
  • Technical: Camera/rendering details. “Wide angle, natural lighting, shallow depth of field.”

    • Combining these: “A calico cat sitting on a windowsill in a sun-drenched Parisian apartment, white curtains billowing, watercolor illustration style, soft edges, muted warm palette, wide angle composition, natural lighting.”

    Use Specific Adjectives, Not Vague Ones

    Vague: “A beautiful landscape”

    Specific: “A misty fjord at dawn, steel-blue water reflecting snow-capped peaks, thin fog layer at the waterline, dramatic sky with pink and orange clouds”

    The specific version gives the model concrete visual anchors. Every adjective should correspond to something visible in the image.

    Control Composition with Photography Terms

    These terms reliably influence composition across all major tools:

    Iterate Systematically

    Do not rewrite your entire prompt when the result is not right. Change one element at a time. If the lighting is wrong, adjust only the lighting terms. If the style is off, swap only the style descriptors. This lets you build a mental model of how each term affects the output.

    Commercial Licensing: What You Can Actually Use

    Licensing is the question that matters most for professional use, and the answer varies dramatically by tool.

    DALL-E 3: OpenAI grants full commercial rights to images you generate, including for products, marketing, and resale. No attribution required.

    Midjourney: Paid subscribers get commercial usage rights. Free tier users do not — images generated on free trials are licensed for non-commercial use only. If your company earns over $1M annually, you must be on the Pro or Mega plan.

    Stable Diffusion: The open-source models (SDXL, SD3) use permissive licenses that allow commercial use. However, fine-tuned community models may have their own license restrictions — always check. Models you fine-tune yourself on your own data are yours to use commercially.

    Flux: Flux.1 Schnell uses Apache 2.0 — fully commercial, no restrictions. Flux.1 Dev is research-only (non-commercial). Flux Pro via the API includes commercial rights with your subscription.

    Ideogram: Paid plans include commercial usage rights. Free tier does not.

    Important caveat: Commercial usage rights from the tool provider do not address copyright questions about the training data. The legal situation around AI-generated images and copyright is still evolving. For high-stakes commercial uses (product packaging, major ad campaigns), consult with a lawyer familiar with AI intellectual property law.

    Integrating Image Generation Into Your Workflow

    For Designers

    Use AI generation as the first step, not the final output. Generate 10-20 variations of a concept, select the strongest direction, then refine in Photoshop or Figma. This collapses the ideation phase from hours to minutes. Midjourney or Flux Pro for initial concepts; Stable Diffusion with ControlNet when you need outputs that match a specific layout.

    For Developers

    Build image generation into your application using APIs. The Stability AI API and Flux API offer REST endpoints that accept a prompt and return an image. For cost-sensitive applications, run Stable Diffusion or Flux Schnell on your own GPU infrastructure — after the hardware cost, generation is essentially free.

    For Marketers

    Establish a prompt library — a documented set of prompts that produce consistent results for your brand. Include your brand colors, preferred styles, and composition guidelines in every prompt. This creates visual consistency across generated assets without needing to brief a designer each time.

    The Bottom Line

    No single AI image generator is best for every use case. Midjourney leads on aesthetic quality. Stable Diffusion and Flux lead on flexibility and cost control. DALL-E 3 leads on accessibility and text rendering. Ideogram leads on typography-heavy images.

    The most effective approach is knowing two tools well: one for quick, high-quality output (Midjourney or Flux Pro) and one for precise control and high-volume work (Stable Diffusion or Flux Schnell). Master the prompting fundamentals — structured descriptions, specific adjectives, photographic terms — and they transfer across every tool. The generator is just the engine; your prompting skill is what steers it.