AI assistants & knowledge
Give people an answer they can use.
Build an assistant around approved information and a clear job. Help a customer find the next step, help a team search its knowledge, or prepare a response for a person to review.
The work
Useful answers need boundaries.
An assistant needs more than a prompt. We define the information it can use, what it may say or do, how it handles uncertainty, and when the conversation should move to a person.
Internal knowledge assistants
Search approved documents, procedures, and reference material with source links. Respect the access boundaries of the audience and give staff a way to flag outdated or incomplete information.
Website & intake assistants
Answer a defined set of questions, collect the details needed for an inquiry, and hand off with context. Keep availability, pricing, and other changing information tied to an approved source.
Drafting & triage
Sort incoming messages and prepare responses or summaries for review. Explicitly separate answering a question from taking an action in a connected system.
Define the finish line
Working means
we can show it.
The acceptance plan is part of the scope. We agree on the checks that matter to your project before implementation begins.
A test set includes answerable questions, missing information, contradictory sources, and requests outside the assistant’s scope.
Users receive only the information their role is allowed to access.
Source updates, human handoff, and an operator’s stop control are tested before release.
Scope & investment
A clear scope. A useful result.
We start with the outcome and the boundaries of the work. Your proposal separates the professional work, any third-party costs, and the support needed after handover.
Quoted to your scope
The fee and payment milestones are agreed before work begins. Changes to the scope are discussed and priced separately.
Our approach to pricingA written scope sets out
- Audience, knowledge sources, and permitted actions
- Evaluation examples and response requirements
- Human handoff and knowledge-maintenance responsibilities
What shapes the quote
- Source quality, permissions, and content volume
- Connected tools and customer channels
- Evaluation requirements, usage, and hosting choice
Ongoing support is scoped to your business: covered systems, planned work, availability, and response expectations are agreed in writing.
Before we begin
A few useful answers.
Can it make up an answer?
AI can produce incorrect information. We design around approved sources, test uncertain and out-of-scope questions, and define when the assistant should decline or hand off. The right review level depends on the consequence of an error.
Can this run locally?
A local deployment can be evaluated against the required model quality, hardware capacity, source systems, and operating responsibilities. Hosted and combined approaches may also fit.
Will it replace our team?
The scope defines the work the assistant helps with and the decisions that remain with people. Useful starting points are finding information, preparing drafts, and gathering context for a handoff.
Who keeps the answers current?
The project identifies the approved source owner and the update process. Ongoing support can include knowledge updates and evaluation against a maintained test set.
Tell us what needs to work.
Share the situation, the systems involved, and what a useful result would look like. We will help define the next step.
Start a conversation