AI Agents and Knowledge Assistants

Assistants that find information in approved business documents and answer with references to the source.

Book a free call Discuss this for your business

How it works · From internal question to sourced answer

STEP BY STEP

Policies
Guides
Past proposals

1
Ask01
A colleague asks in Teams or Slack
“Refund policy?”
2
Search02
Relevant passages retrieved
3 passages found
3
Check03
Permissions and dates verified
Access: allowed
4
Answer04
Concise answer with references
2 sources cited
5
Escalate05
Unclear questions go to an expert
Expert notified
Human checkpoint

Out

Answer with sources
Feedback
Gap report

See how this would run on your own process, with your tools.

Book a free call

The problem

What this solves.

Policies, guides, and notes are spread across folders and tools. People spend time searching or asking colleagues the same questions.

What changes

Answers in seconds, with the source attached
Consistent access to current guidance
Fewer interruptions for senior staff
Visibility of gaps in your documentation

Use cases

What this looks like in practice.

Typical examples. Every build is scoped to your own process and systems.

Policy assistant

Staff ask questions about HR, finance, or operating procedures in Slack or Teams and get an answer with a link to the exact section it came from.

Proposal drafting

Past proposals and case notes are indexed so a first draft can be assembled from approved material, then edited by the team.

Meeting summaries

Recorded calls are summarised into decisions and actions, which are filed against the right client or project.

Scope

What we can build.

Capabilities

Internal knowledge assistants
Document search and question answering
FAQ and support assistants
Structured request intake
AI-assisted drafting and summarisation
Human-reviewed response workflows

How it works

We agree which sources the assistant may use and who may see what, index that content, and have the assistant answer from retrieved passages with references back to the source.

Tools we commonly work with

Slack
Microsoft Teams
Google Drive
SharePoint
Notion
Confluence
OpenAI
Anthropic Claude
Google Gemini

Limits and human review

An assistant is only as useful as the information, permissions, and accuracy behind it. It should say when it cannot answer and point to a person. It does not make decisions on its own.

Delivery

From first call to working system.

01
Discover
Days
02
Map
1–2 weeks
03
Design
Days
04
Build and test
2–6 weeks
05
Improve
Ongoing

You receive

Indexed, permission-aware knowledge base
Assistant in the agreed channel
Evaluation test set and results
Content maintenance guide

The Shapes Method

Workflows

Workflows in this service.

Each one has its own page with a diagram of how it runs.

Questions

Common questions.

Is our data used to train AI models?

We use providers and settings that do not train on your data by default, and confirm the data handling terms of each service before it is used.

Where are our documents stored?

In a vector database in your own account, or one we set up for you. Access follows the permissions you already have.

Can an agent take actions, not just answer?

Yes. Tool-using agents can create records, send drafts, or update systems, with approval steps wherever the impact warrants it.

Next service · 13Integrations and Custom Internal Tools

Next step

What would you like to improve?

Tell us what is taking too much time, what is difficult to manage, or what you would like your systems to do differently. You do not need a technical specification to start.

Book a free call Send a message