Scoped roles
Only the right people can act in the right places.
The platform for AI workplace apps
Governed AI agents that move real business work forward
AI WorkBook is a workspace for designing role-specific agents with approved context, connected tools, and a clear path from request to review, approval, and action.
Scoped roles
Approved tools
Human oversight
Audit trails
Request
AI agent
Review
Approval
Built-in governance
Approved tools
Agents only use connected tools you approve.
Human oversight
Critical decisions stay in human hands.
Audit trails
Every action is recorded with full traceability.
Incoming work
Contract review request
A supplier asks to change payment terms from 30 to 90 days.
Owner: ProcurementAssigned agents
Analysis Agent
Finds policy and history
Completed
Review Agent
Flags commercial risk
In progress
Output Agent
Drafts the reply
Pending
Next step
Manager approval required
The recommendation, evidence, and draft response are saved together.
Status: Decision pack ready
3 sources checked
Risk flagged
Draft created
Approval logged
Built for governance and trust
View audit trail
Every step is traceable. Every decision is accountable.
What makes an AI WorkBook agent different
More than a chatbot. More controlled than open-ended automation.
AI WorkBook agents are special because the intelligence sits inside a designed work pattern: a role, trusted context, approved actions, and a clear handoff to a person, system, or next agent.
They understand a role
Each agent is shaped around a specific job in a process, not a general chat window.
- Clear responsibility
- Defined output
- Known handoff
They use approved context
The agent can work with the policies, records, documents, and tools it is allowed to use.
- Trusted sources
- Scoped permissions
- Connected tools
They stay governable
Important decisions can be routed for review, saved with evidence, and checked later.
- Human approval
- Evidence packs
- Audit trail
How they work
One request becomes a controlled chain of work.
Supplier message
Can we move to 90-day payment terms?
AI WorkBook turns the request into checks, evidence, a recommendation, and an approval step.
The agent identifies the term change, supplier, contract owner, and decision needed.
It reads the contract, finance policy, supplier record, and previous decisions.
It explains the risk, suggests the position, and drafts a supplier reply.
It routes the evidence pack to the right person before anything is sent.
Sources used: contract, policy, supplier record
Decision owner: procurement manager
Action: waiting for approval
Equipped for the job
An agent is not just prompted. It is set up to work safely.
Before an AI WorkBook agent can help, it is given the right role, knowledge, tools, limits, and handoff path. That setup is what turns AI from a chat box into a useful part of a business process.
Knowledge and memory
Records, documents, policies, history, and useful notes from previous work.
Approved tools and actions
Search, checks, APIs, parsers, validators, generators, and small approved actions.
AI WorkBook Agent
Researcher Agent
- Role
- Finds the right evidence
- Current task
- Check supplier payment terms
- Goal
- Deliver a clear, well-sourced recommendation
Workflow context
Stage, permissions, owner, progress, decisions, and where the work should go next.
Outputs and actions
Draft a response, create a summary, update a record, or trigger the next step.
Policies
Security
Human approval
Audit trail
What agents can be equipped to do
The practical jobs agents do inside the workflow.
AI WorkBook agents are assembled from proven workplace capabilities, then configured around the process, people, and controls your team uses.
Capture the request
Turn emails, forms, notes, or tickets into a clear piece of work.
Find the right context
Pull in approved policies, records, history, and documents.
Check the risk
Compare the request with rules, thresholds, and previous decisions.
Draft the response
Create the summary, reply, report, plan, or next-step recommendation.
Ask for approval
Route the decision to the right person with the evidence attached.
Log the decision
Record what was used, who approved it, and what changed.
Anatomy of an agent
Each agent combines intelligence, context, tools, and governance.
AI WorkBook lets teams create agents that are useful because they are specific. Each one has a job to do, trusted context to use, approved actions it can take, and controls around important decisions.
Job and outcome
The role it plays, the decision it supports, and the result it should produce.
Approved context
The documents, records, policies, customer context, and prior decisions it is allowed to use.
Tools and actions
The checks, calculations, drafts, handoffs, and system updates it can carry out.
Controls and review
The permissions, approval moments, quality checks, and logs that keep the work accountable.
Specialist agents, organised by role.
The same workflow can be split across specialist agents, so each part of the process has the right context, rules, and owner.
01
Strategy Agent
Sets the policy position and decision threshold.
02
Coordination Agent
Finds the owner and keeps the request moving.
03
Analysis Agent
Checks contracts, policy, and past decisions.
04
Review Agent
Flags risks that need human judgement.
05
Output Agent
Creates the draft response and evidence pack.
1
Evidence attached
Contract, policy, and supplier record
2
Approval needed
Legal review before response
3
Decision owner
Procurement manager assigned
4
Review trail
Every source and action recorded
Governed by design
Agents are powerful because every action has a boundary.
AI WorkBook agents are designed to work inside clear limits: what they can read, what they can draft, when they need approval, and what must be saved for review.
Approved sources
Clear permissions
Approval moments
Saved evidence
Risk checks
System handoffs
Continue exploring
See how governed agents fit inside the wider AI WorkBook platform.
Return to the platform page for the bigger story: workflow apps, product examples, reusable building blocks, and why different teams care about AI WorkBook.