botzone.ai
Back to Blog
AI AgentsTeamMulti-AgentProduct

Meet the Line-Up: Seven Products, One Idea

Stephen Keegan||6 min read

One idea, seven products

Most AI companies ship one assistant and keep bolting features onto it. We went the other way. Botzone is a line-up of separate products, each one narrow, each one aimed at a job somebody already does badly today.

They do not share a login. They do not upsell each other. You can use one and ignore the rest, and most people do. What they share is a single conviction: an AI product should ship the work, not a summary of the work.

Here is the whole line-up as it stands today, what each one actually does, and how far along it is.


Anna: the assistant for busy trades and small businesses

Status: Beta

Anna answers your business's WhatsApp, day and night. She reads text, photos, and voice notes in any language, works out what the customer actually needs, and hands you the whole enquiry as a single job card in your inbox.

She is built for the electrician up a ladder and the plumber under a sink. The problem is not that these businesses are bad at customer service. It is that you cannot answer the phone with your hands full, and by the time you are back in the van the customer has rung the next fella.

The design constraint that matters most: Anna never books anyone in and never invents a price. She defers anything unusual back to you. She is built and tested never to state anything that is not already in your own knowledge base, which is assembled from your website, your reviews, and your trade registration.

What she does:

  • Answers 24/7 on WhatsApp - text, photos, voice notes, any language
  • Sends you a job card with the urgency, the issue, the area, and the full transcript
  • Defers anything unusual instead of guessing
  • €49 a month, first month free. anna.botzone.ai


    Sage: meeting intelligence

    Status: Live

    Sage is the oldest product in the line-up and still the busiest. She joins your Google Meet and Microsoft Teams calls as a participant, announces herself, and listens.

    She is not a recorder. While the call runs she is classifying what is said into decisions, action items, and open questions, each one timestamped and attributed. Ask her out loud who is missing, or what has been decided so far, and she answers in seconds.

    Sixty seconds after the call ends, a structured document lands. Decisions at the top, action items with owners, open questions that still need resolving. Not a transcript dump.

    What she does:

  • Joins Meet and Teams calls and announces herself
  • Extracts decisions and action items live, with ownership
  • Delivers a structured recap 60 seconds after the call ends
  • sage.botzone.ai


    Cole: tech help that fixes it

    Status: Beta. Desktop only, because it needs screen sharing.

    Cole watches your screen, talks you through the fix, and draws arrows where you need to click. It is the experience of FaceTiming the family member who is good at computers, except it is always available and never sighs.

    The important part is what Cole does not do: it never takes over your machine. Cole points, you click. There is no install and no remote-desktop agent. It runs in the browser.

    What it does:

  • Sees your screen, with no install
  • Talks you through the fix out loud, with on-screen arrows
  • Leaves you in control of your own machine
  • cole.botzone.ai


    Otto: a personal AI mentor

    Status: Private beta, waitlist only.

    Otto is the newest and the least finished, and it is worth being precise about that. It is a macOS desktop app in a single-user pilot, not a hosted product you can sign up for today.

    What works: Otto captures your Zoom and Google Meet calls, transcribes them on your own machine using a local Whisper model, and reads your Gmail, Google Calendar, and Slack to build a picture of who you talk to and what you committed to. Thirty minutes before a meeting it sends you a short prep note. Ask it a question and it answers with citations back to the source.

    The whole design bet is that this class of product should run locally. Audio never leaves the machine. Transcription and embeddings both run on-device.

    What it does today:

  • Captures Zoom and Meet calls, with a cancel window before anything is kept
  • Transcribes on-device - the audio never leaves your Mac
  • Sends a short prep note before each meeting, and answers questions with sources
  • otto.botzone.ai


    Cost: LLM cost attribution

    Status: Beta

    Cost is the developer tool in the line-up. It wraps your Anthropic, OpenAI, and Gemini clients with one call, and attributes every euro of spend to a route or a feature.

    Then it does the part nobody else does. When it spots a place you could run a cheaper model, it does not just tell you. It shadow-runs the cheaper model against your real traffic, has an independent judge grade every answer against the original, and only surfaces the switch once it has passed. You keep the saving.

    What it does:

  • Drop-in SDK for TypeScript and Python, two lines to install
  • Real spend in euros, mapped to your routes and features
  • Downgrade recommendations that are eval-gated before you ever see them
  • Free under €500 a month of AI spend. cost.botzone.ai


    Lex: EU AI Act compliance, signed off

    Status: Live

    Lex maps every AI system you run to its EU AI Act risk tier and its GDPR duties, then produces a gap report and a roadmap. A Dublin-based AI practitioner reviews and personally signs every one.

    This matters more than it did six months ago. Article 50 of the Act, the transparency obligation that requires a chatbot to disclose it is a machine at first contact, becomes enforceable on 2 August 2026.

    What it does:

  • Classifies every system against the Act, with the Articles that apply
  • A PhD AI practitioner reviews and signs each report
  • Optional quarterly monitoring as the regulation and your systems change
  • lex.botzone.ai


    Watch: agentic-AI security, continuously checked

    Status: Live

    Watch is for teams shipping AI agents of their own. It snapshots an agent's exposure, grades it A to F, and keeps re-checking as your prompts, tools, and models change.

    The scoring methodology is open source, on the theory that a security grade you cannot audit is not worth much.

    What it does:

  • Scan an agent, get an A-F posture grade in minutes
  • Re-checked continuously as your setup changes
  • Transparent, open-source scoring
  • watch.botzone.ai


    Why separate products, not one platform

    The temptation is to build one assistant that does everything, put it behind one login, and call it a platform. We tried a version of that early on. It does not work, for the same reason a general-purpose employee is less effective than a specialist.

    Each of these jobs has a different user, a different definition of done, and a different failure mode. An electrician who misses a WhatsApp message and a platform engineer who cannot explain an Anthropic invoice have nothing in common except that both problems can be solved with a language model. Forcing them into one product would make both worse.

    So each one gets its own product, its own subdomain, and its own pricing. Every one is free to try. If only one of them is useful to you, that is a complete outcome, not a failed cross-sell.

    What changed since this post first ran

    An earlier version of this article introduced three agents: Sage, Cara, and Dash. Cara and Dash were retired. Cara's work was folded into Anna, which turned out to be a far better fit for the customers who actually wanted it, and Dash was cut entirely because standup facilitation was a thinner problem than it looked.

    The line-up above is the current one. When it changes again, this post changes with it.

    Share