CREATE
Start with a job, not a model.
Tell TAZM what the bot should do. Define the job, the outcome, and the boundaries.
In the prototype
The TAZM Bot Factory
Build a new bot.
Or bring the one you already have.
Create, train, control, test, certify, store, and deploy your AI workforce with TAZM.
Your bot. Your knowledge. Your rules. TAZM provides the control layer.
The Bot Factory is in development: you cannot build or bring a bot yet. You can create your TAZM account now — this page will say when the Bot Factory opens.
Illustration
Refund Helper
Answers refund questions from your policy
JobDefined
KnowledgeRefund policy, help centre
Read customer recordsAllowed
Issue a refund up to $50Needs approval
Delete a recordForbidden
Evaluation12 cases written
Where it stands today. The Bot Factory is in prototype. Describing the job, setting
authority and writing test cases work in the prototype. Importing, training, certification,
storage and deployment are designed and coming — each is marked below.
What you can do
Everything a bot needs to go from idea to trusted worker.
IMPORT
Bring an existing bot.
Already built a bot somewhere else? Bring its instructions, knowledge, workflows, tools, and examples into TAZM.
Coming
TRAIN
Improve how it performs.
Teach with examples, corrections, documents, demonstrations, and real work.
Coming
CONTROL
Define its authority.
Choose what the bot may do, what requires approval, and what is forbidden.
In the prototype
TEST
Evaluate before deployment.
Run test cases, edge cases, tool-use checks, and policy evaluations before the bot is trusted.
Writing tests: in the prototype · Running them: coming
CERTIFY
Prove it meets your standard.
Certification means the bot passed evaluation and policy gates — not that somebody clicked Publish.
Coming
STORE
Keep your bots organized.
Store bots, versions, knowledge, policies, evaluations, and deployment history in one place.
Coming
DEPLOY
Use it where the work happens.
Deploy through approved channels when the bot is ready.
Coming
The Bot Factory
Your AI workforce starts in the Bot Factory.
The Bot Factory is where companies create and improve reusable AI workers — each one built around a job, taught your way of working, held to your rules, and proven before it is trusted.
1
Describe the job
In plain words
2
Add knowledge
Your documents and policies
3
Connect tools
Where it does the work
4
Set authority
What it may do
5
Train
Examples and corrections
6
Test
Cases it must pass
7
Certify
A gate on evidence
8
Deploy
Where the work is
In the prototype today: describing the job, setting authority and writing tests. The other stations are designed and coming.
Bring your existing bot
Already have a bot? Don’t rebuild it.
Import what you already have and put TAZM around it: the same bot, now with a defined job, your rules, real evaluations and a record of what it did.
Existing chatbotInternal support bot
Sales botResearch bot
Automation botCoding agent
Customer service agent
Import is coming: no bot can be imported yet, and no integration with another platform exists today. It is designed to start from what any bot has — its instructions, knowledge and examples — with tools and workflows following as connectors are built.
1Your existing bot
2ImportComing
3TrainComing
4Add authorityIn the prototype
5EvaluateComing
6CertifyComing
7DeployComing
Illustration
BotVersionState
Refund Helperv3Awaiting certification
Invoice Checkerv7Needs work
Sales Researcherv2In testing
Policy Answererv1Draft
An illustration of the library to come. None of these bots exists.
Store and manage your bots
One home for every bot you build.
Your bots should not live scattered across prompts, laptops, repos, and provider dashboards. In TAZM, each one keeps everything that makes it what it is:
- Bot identity
- Job definition
- Instructions
- Knowledge
- Tools
- Authority
- Training history
- Evaluations
- Certifications
- Versions
- Deployment state
- Evidence and activity
Coming The library is designed. In the prototype, a bot is a draft kept in your browser.
Authority
Decide what your bot can do before it does it.
Every action is one of three. Anything you have not listed is forbidden.
Allowed
Read CRM
Summarize records
Requires approval
Send email
Issue refund
Modify production
Forbidden
Delete database
Access payroll
Export secrets
The bot
asks to act
asks to act
TAZM
checks the action against your rules
checks the action against your rules
The action
goes, waits for a person, or is refused
goes, waits for a person, or is refused
TAZM sits between the bot and the action. You can write this policy in the prototype today; enforcing it arrives with the runtime.
Training
Teach the bot how your business actually works.
Training in TAZM teaches the job — your documents, examples and corrections shape how the bot does the work. It does not change the model underneath.
DocumentsExamplesCorrections
DemonstrationsPoliciesExpected outputs
Train
Evaluate
Correct
Re-test
… and round again, until it meets the bar you set.
Coming No bot is trained yet. In the prototype you can write the instructions and examples it will learn from.
Certification
Don’t deploy because it looks ready. Prove it.
Job completerequired
Knowledge configuredrequired
Authority configuredrequired
Tests passedrun, not assumed
Policy checks passedrun, not assumed
No critical violationsobserved, not assumed
CERTIFIED — every requirement met, on evidence.
NEEDS WORK — something is missing or unproven, and the gate says what.
REFUSED — it broke a rule it must never break.
Coming No bot has been certified. In the prototype the gate shows what is missing; it cannot issue a certificate.
Use cases
One factory. Every kind of job.
Customer Support
Answers from your policies and hands off what it cannot.
Sales Research
Researches accounts and briefs your reps.
Operations
Runs routine checks and reports the exceptions.
Finance
Checks invoices and prepares them for approval.
Security Review
Reviews changes against your security rules.
Recruiting
Screens applications against the role you define.
Real Estate
Qualifies enquiries and proposes viewings.
Research
Turns sources into sourced notes.
Internal Knowledge
Answers staff questions with the document it came from.
Developer Agents
Works in your repositories under your rules.
Provider independence
Your bot is bigger than the model underneath it.
TAZM is the control layer. Providers are implementation choices. The job, the knowledge, the rules and the evidence are yours, and they stay with the bot.
Build your first bot — or bring the one you already have.
The Bot Factory is in development. Creating an account does not make any bot run.