Operations
Knowledge management
Answers are buried in documents, inboxes and people’s heads.
The problem
Why it happens.
Knowledge builds up faster than anyone organises it. A policy is written, then updated in an email, then explained differently in a training session, and each version lives somewhere else. Shared drives fill with drafts and copies, and a search returns a dozen files with similar names.
So people stop looking and ask a colleague. That works while the experienced people are available, but it interrupts their own work, and answers drift as they pass from person to person. When someone leaves, part of what they knew leaves with them.
Signs you have this problem
- New starters spend their first weeks asking colleagues where things are.
- The same question gets a different answer depending on who you ask.
- There are three versions of the same policy and nobody is sure which is current.
- One long-serving person is interrupted all day with “how do we…?” questions.
- Customers are told one thing on the phone and another by email.
The answer
In plain terms.
Every business has knowledge that lives in a handful of places: a shared drive, old email threads, a policy document nobody can find and the memory of the person who has been there longest. Knowledge management with AI brings the approved sources together and puts an agent in front of them. Staff and customers ask in plain words and get an answer drawn from your own material, with the source shown.
What changes
One reliable answer, wherever the question is asked.
How it works
- 01
Gather the sources
We agree which documents, policies and guides are the approved source of truth, and leave out anything out of date.
- 02
Set who can see what
Internal guidance stays internal. Customer-facing answers draw only on material you are happy for customers to read.
- 03
Build the agent
The agent answers from those sources, shows where an answer came from and says when it does not know.
- 04
Keep it current
A named person owns the sources, and updates flow through to the agent when documents change.
A typical rollout
From first call to live.
Timings are a rough guide. They depend on your systems, your availability and what the discovery work finds.
- 01 · First week or two
Audit the sources
We gather the documents people rely on, agree which are current and approved, and set aside anything out of date.
- 02 · The next couple of weeks
Set access and build
Access rules are agreed for each audience and the agent is built on the approved library, starting with internal use.
- 03 · Around the end of the first month
Test with staff
A small group asks real questions, checks the sources and flags gaps before the agent is opened to the wider team.
- 04 · Ongoing
Extend and maintain
Customer-facing use follows once internal answers are reliable, and the owner keeps the library current.
What you get
- An agreed library of approved sources
- An AI agent that answers from that library
- Access rules separating internal and customer-facing knowledge
- An owner and routine for keeping the sources up to date
What it connects to
- Shared drives and document stores
- Intranet or internal wiki
- Website chat
- Helpdesk or CRM
Who it suits
- Teams where new starters rely on asking colleagues
- Businesses with policies or procedures that change often
- Customer-facing teams who give slightly different answers to the same question
A typical example
How it plays out.
A lettings agency’s staff answer tenant and landlord questions from a mix of policy documents, templates and past emails. With a knowledge agent in place, a new member of staff can ask how the deposit process works and get the agency’s own answer with a link to the policy. Tenants asking the same thing through the website get the customer-facing version from the same source.
A generic scenario to show how the work runs. It does not describe a specific client.
Measuring it
How you know it is working.
- Questions answered with a source
- An answer with a source can be checked. This shows how much of what people ask is covered by the library.
- Questions the agent could not answer
- These point to missing or unclear documents, which makes them the owner’s to-do list.
- Interruptions to experienced staff
- Ask the people who used to field the questions whether fewer reach them. It is a simple check that the agent is being used.
- How new starters get on
- New starters are often the heaviest users. Their feedback shows whether the library covers what they need in their first weeks.
- When each source was last reviewed
- Tracking review dates stops old guidance from being served as current.
Risks and safeguards
What could go wrong, and what stops it.
Risk
The agent gives an out-of-date answer.
Safeguard
Only approved sources are used, each has a review date and a named owner, and replacing a document replaces what the agent draws on.
Risk
Internal information reaches customers.
Safeguard
Separate access rules decide which sources each audience can draw on, and customer-facing answers come only from material marked for customers.
Risk
The agent fills a gap with a guess.
Safeguard
It is set up to say when it cannot find an answer and to point the person to the right contact.
Risk
Staff treat the agent as the final word on a judgement call.
Safeguard
The agent answers from policy and shows the source. Decisions that need judgement or authority, such as exceptions to a policy, stay with a named person, and the agent says so.
Risk
Sensitive documents leave your control.
Safeguard
Personal and confidential material can be kept out of the library, or the agent can run on a private model. We agree the approach against your UK GDPR obligations before any documents are loaded.
Questions
What people ask about this.
Where does the agent get its answers?
Only from the sources you approve. It shows which document an answer came from and says when it cannot find one.
Is our internal information kept private?
Access rules decide which sources each audience can draw on. For sensitive material, the agent can run on a private model. See AI infrastructure and private models.
What happens when a policy changes?
The owner updates the source document and the agent uses the new version. Keeping sources current is part of the setup.
What kinds of documents can it use?
Policies, procedures, guides, templates, FAQs and similar written material in common formats. Part of the setup is deciding which of them are current enough to be a source.
Can staff and customers use the same agent?
They can share one library through different access rules. Staff can draw on internal guidance, while customers only see material approved for them.
How do we know the answers are right?
Each answer shows its source, so anyone can check it. Before launch we test the agent against real questions, and the owner reviews the questions it could not answer.
Next step
Start with the free AI Opportunity Assessment.
A short, no-obligation conversation about where your business is losing enquiries, hours and revenue. You do not need to choose a package first, and whatever we find carries straight into your audit, so your first paid day starts from evidence.