AI Products & Integrations
Build an AI product that does a real job.
From a customer-facing feature to an internal agent or a new AI-native product, we design, build and integrate the smallest useful version—grounded in approved data, measurable quality and clear human control.
Scope an AI productStart with the job
Use AI where it removes a real constraint.
The useful question is not which model to buy. It is where a customer or team is losing time, context or confidence—and what a controlled product could make easier.
People cannot find the right answer, option or next step.
A bounded search, assistance or recommendation feature can help when it works from approved information and has a clear path when confidence is low.
A valuable feature is trapped in the backlog.
A focused integration can test one customer job inside the current product without turning the whole platform into an AI experiment.
The team repeats document, research or triage work by hand.
A reviewed workflow or agent can prepare, classify or route work while a person keeps authority over material decisions.
The idea is promising, but the smallest credible release is unclear.
A prototype or technical spike can test the riskiest assumption before a broader product build, platform commitment or operating cost.
Four practical needs
Choose the smallest useful form of AI.
The service covers website visibility, existing-product integrations, reviewed workflows and new products. Conventional custom software remains available when AI is not the right mechanism.
AI visibility for websites
Clarify entities, source facts, page relationships and actions so AI systems can interpret the website more accurately. This improves machine readability; it does not guarantee inclusion in an external answer.
AI features in existing products
Add a defined search, assistance, summarisation, extraction or recommendation job to a product people already use, with the current accounts, permissions and data boundary respected.
AI workflows or agents with human review
Support repeated research, document, service or operations work with approved sources, evaluation cases, logged decisions, review points and a clear human fallback.
Net-new AI products and MVPs
Shape an AI-native product, customer application, internal platform or conventional software product around the smallest production release that performs a valuable job.
A bounded buying path
Blueprint → Pilot → Build → Care
Each step earns the next one. The Blueprint is the paid first engagement; larger implementation and ongoing care are priced only after the workflow, data, risk and integration boundary are understood.
- 01 / BLUEPRINTStarting from $2,000
AI Product Blueprint
A paid, fixed-scope first engagement that turns the user need and workflow into an informed build decision.
Typically 5–10 business days, subject to confirmed scope and scheduling.- User and workflow map
- Approved data and integration map
- Risk and human-control map
- Prototype or technical spike
- MVP scope and build estimate
The final Blueprint fee is agreed before work begins. It may increase for multiple workflows, sensitive data, extensive integrations or unusual technical validation.
- 02 / PILOTQuoted after the Blueprint
AI Pilot Build
One useful production slice that proves the core workflow under real operating conditions before a larger build.
Timing is confirmed against the approved pilot scope.- Authentication, data and integrations where required
- Evaluation cases and acceptance criteria
- Guardrails, monitoring and human fallback
- Production release evidence and handover
There is no universal pilot price. The quote follows the approved Blueprint and names the exact production slice.
- 03 / BUILDQuoted from approved discovery
Custom AI Product Build
A broader customer-facing product, internal platform, AI-native MVP or integration into an existing system.
Delivery stages and commercial milestones follow the approved scope.- Product and interface implementation
- Application, data and integration architecture
- Evaluation, safety and operational controls
- Release, documentation and technical handover
The build is quoted only from approved discovery output. Conventional application work stays in scope when it serves the product better than an AI feature.
- 04 / CARESeparately agreed after launch
AI Product Care
Post-launch evaluation and improvement for a product that already has a clear owner, operating boundary and release history.
Cadence and response boundaries are agreed separately.- Evaluation review and regression checks
- Model or prompt changes
- Cost, latency and reliability work
- Incident review and iterative improvements
AI Product Care is not included by default. It is a separate engagement quoted against the live product and required cadence.
Delivery controls
Make quality, risk and authority visible.
AI output can vary. The delivery system must therefore show what the product is allowed to use, how quality is tested, when a person decides and what happens when the system should not continue.
Approved inputs only
Name the source material, permissions, retention boundary and integrations before the product relies on them.
Test the job, including failure cases
Use representative evaluation cases, acceptance criteria and recorded limitations instead of treating a polished demonstration as proof of production quality.
Keep material decisions with the right person
Define review points, escalation, override and fallback so automation never quietly expands beyond its approved authority.
Monitor cost, speed and reliability
Track usage, errors, latency and meaningful quality signals so the team can investigate a problem and decide whether a change is worth releasing.
Ownership and boundaries
Leave with the product, its controls and a documented handover.
The proposal records what becomes client-owned, what is licensed from elsewhere and which accounts remain under third-party terms. The handover is designed to keep the operating knowledge beside the code.
The agreed handover can include
- Custom source code and interface assets created in scope
- Prototype or technical-spike artifacts
- Evaluation cases, acceptance criteria and known limitations
- Architecture, integration, deployment and operating documentation
- Agreed account access, configuration records and release evidence
External and separately agreed boundaries
- Client data and business assets remain the client’s; pre-existing and third-party software keeps its own licence.
- Model, API and infrastructure usage charges are separate unless the proposal explicitly includes them.
- App stores, model providers and other externally controlled services retain their own approval, availability and commercial terms.
- Sensitive or high-stakes use may require additional legal, security, privacy or domain-specialist review before a build can be accepted.
- No model output, platform approval or external-service availability is guaranteed.
Questions before scoping an AI product
Do we need a finished specification first?
No. The AI Product Blueprint exists to turn a user need, workflow and known constraints into a prototype or technical spike, an MVP scope and a build estimate.
Can you add AI to a product we already use?
Yes, when there is a defined job, approved data, workable account access and a safe path when the feature is uncertain or unavailable. The first version should stay narrower than the whole product.
Do you still build conventional custom software?
Yes. Customer applications, portals, internal platforms, mobile or desktop products and integrations remain within this service. AI is used only when it improves the product’s real job.
What can raise the Blueprint above $2,000?
Multiple workflows, sensitive data, extensive integrations or unusual technical validation can increase the fixed fee. The final amount and scope are agreed before work begins.
Can you guarantee accurate AI output?
No universal accuracy promise is credible. The engagement defines representative evaluation cases, acceptable performance, known failure modes, review points and fallback for the specific job.
Are model and API charges included?
Not by default. Third-party model, API, infrastructure and platform charges keep their own terms and are identified separately in the proposal.
Is post-launch care included?
No. AI Product Care is a separately agreed engagement for evaluation review, model or prompt changes, cost and latency work, incident review and iterative improvement.
Start with a bounded first decision
Define the job, the approved data and the human control before the build expands.
Use the private project brief to describe the user, workflow, existing product or system, and the result that should become easier.
Scope an AI productAI Product Blueprint starts from $2,000. The final fee is agreed before work begins.