AI chatbots
An AI chatbot answers in conversation — on a website, in a help center, in Slack — grounded in an approved body of your client’s real content, with a defined scope of what it must decline.
For agencies whose clients field the same questions all day: support queues, policy lookups, pre-sales questions a page never quite answers.
- Fixed-price / build
- $3,000–$8,000
- Pilot to proof
- 2 weeks
- Defect warranty
- 90 days
Where the engineering actually is.
The engineering that matters in a chatbot isn’t the conversation — models handle that. It’s the grounding. The assistant answers from an approved set of sources: the client’s help center, docs, policies. Ask it something outside that set, and the correct behavior is to say so and hand off, not to guess.
A support assistant that invents a refund policy is worse than no assistant — the refusal scope, the escalation path, and the transcript trail are the deliverable, not garnish on it.
Every assistant ships with analytics your agency can put in front of the client: what was asked, what was answered, what was handed off, and where the content has gaps worth fixing.
If the assistant needs to take actions rather than give answers — update a record, file a ticket — that’s an agent build. Read about AI agents.
Where a chatbot earns its keep.
Support deflection
The client’s team answers the same twenty questions on repeat. A grounded assistant resolves the repeatable ones and hands the rest to a human with full context — so the humans work the queue that actually needs them.
Internal knowledge
Policies, procedures, project history — knowledge that exists but lives in forty documents. An internal bot in Slack or the intranet answers with the source attached, so every answer is checkable.
Lead capture
A site visitor with a real question either finds the answer or leaves. An assistant answers from the client’s own content, collects contact details when intent is clear, and routes the conversation to a human or a calendar.
Example chatbot builds.
Illustrative examples — the shape of work this line covers, not client case studies. Builds are quoted fixed-price from the published range ($3,000–$8,000 / build) before work starts.
Help-center support assistant
Grounded in the client’s help center; answers with linked sources, hands off to a human with the full transcript, and reports deflection monthly.
Internal policy assistant
Lives in the client’s Slack; answers HR and operations questions from the approved policy set, and says plainly when a policy doesn’t cover the question.
Website concierge
Answers product and pricing questions from the client’s own pages, and turns real intent into a booked call or a captured contact — without pretending to be a person.
How we build chatbots.
Grounding first: we inventory the content the assistant may answer from, and the spec defines the refusal scope — the questions it must decline and where those go instead. These assistants are grounded in approved content, not fine-tuned on it, so every answer stays source-linked and updatable.
Before launch, QA runs the assistant against a written set of must-answer and must-refuse cases from the signed spec. After launch, content changes are cheap by design: update the source, not the software.
The mechanics are the same on every line we ship:
- 01Pilot first — Every relationship starts with one real project: $1,000 flat, two weeks, fully credited toward month one of any retainer.
- 02Scoped in writing — Scope, milestones, and — on fixed-price work — the price, signed before work starts. The spec is the contract QA runs against.
- 03One point of contact — Javid scopes every project and answers inside four business hours; every active engagement gets a written status each Friday.
- 04Founder QA — Every deliverable is checked against the signed spec before you see it — behavior, edge cases, and security.
- 05White-label handover — Repos, credentials, and docs transfer to accounts you or your client control. Nothing we ship carries our name.
- 0690-day warranty — If a delivered build doesn’t match its signed spec, we fix it free for 90 days after handover.
Technologies
Defaults, not dogma — where your client already has a stack, we build in it.
- OpenAIModels
- AnthropicModels
- SupabaseRetrieval store
- Next.jsSite embeds
- TypeScriptIntegration code
- n8nHandoff workflows
Questions about AI chatbots.
How do you stop it from making things up?
Grounding plus refusal scope. The assistant answers only from approved sources and links them; outside that set, it declines and hands off. The must-answer and must-refuse cases are written into the spec and tested before launch — and if the assistant gets one of those specified cases wrong, that’s a defect the 90-day warranty covers.
Where can it live?
The client’s website, help center, Slack, or an internal tool — wherever the audience already is. The same grounded core can serve more than one surface.
What happens when the client’s content changes?
The assistant reads from sources, not from a frozen training run, so updating the help center updates the answers. Re-testing the must-refuse set after a big content change is a care-plan task or a small scoped one — your choice.
Do visitors know it’s a bot?
Yes. Assistants we ship identify themselves as automated. It’s your client’s brand on the widget — a bot caught pretending to be a person costs them more than it ever saves.
Related service lines.
AI agents
Agents that complete multi-step work — document processing, triage, research — with human gates where it matters.
Voice AI
Agents that hold a phone call — answering, qualifying, booking — with a human transfer path and a transcript for every call.
Maintenance & support
A per-site care plan that keeps shipped systems running: monitoring, updates, small changes, incident response — under your brand.
Start with one real build.
Scope your first project as the pilot — $1,000 flat, two weeks, fully credited toward month one of any retainer. If the work isn’t retainer-quality, you keep the deliverable and walk.