Need someone to actually build your AI solution, not just advise?

Sounds familiar?

• "We paid for an AI strategy. It's a PDF. Nothing runs."
• "The agency built it. Then they left. Now it breaks and nobody here can fix it."
• "I don't need another workshop. I need someone to connect this to our CRM and make it work."
• "We're a 20-person company. The big consultancies won't return our calls and the freelancers can't scale."
• "I want one team responsible for the whole thing, start to finish."

Consultant, implementation partner, or agency — which do you need?

The three are sold with overlapping language and do very different jobs. The fastest way to tell them apart is to ask who is accountable when the thing does not work at 2 a.m.

Our full comparison of AI consultant vs implementation partner vs agency goes deeper, including the questions to ask each one before you sign.

What an implementation engagement includes

• Diagnosis — a short audit of the workflow the system will touch, so we build against how the work actually happens, not the org chart's version.
• Design — data flow, model choice, where the human stays in the loop, failure modes and what happens in each.
• Build — the system itself, in a mainstream stack your next engineer will recognise.
• Integration — with your CRM, phone system, document store, billing and calendar. This is usually most of the work and most of the value.
• Testing — around the paths that matter: the happy path, the bad input, the API outage, the duplicate record.
• Deployment — environment separation, secrets management, CI, rollback.
• Monitoring — you find out it failed before your customer does.
• Documentation and handover — a runbook, and the accounts in your name.
• Run (optional) — we keep operating and improving it on a retainer, or you take it in-house.

What we build most often

• Intake and qualification systems — inbound leads, referrals and forms read, scored and routed.
• Document processing — contracts, claims, invoices and onboarding paperwork extracted and filed.
• Retrieval systems over your own documents (RAG) — answers grounded in your policies, contracts and records, with citations.
• AI agents for back-office work — multi-step tasks across several tools, with approval gates.
• Voice agents for inbound calls — scheduling, intake and FAQs, with a warm hand-off to a person.
• CRM and follow-up automation — the sequence that stops leaking revenue on day three.
• Finishing and productionising AI apps founders have built themselves — see vibe-coding rescue.

How we work

1. Diagnostic call — twenty minutes on the operation and the outcome you need. We say whether it is worth building and roughly what it would take.
2. Audit and design — a fixed-scope engagement that ends with a costed plan and a design you keep, whoever builds it.
3. Build and integrate — our engineers build the system against that design. First system in production in two to four weeks for most engagements; larger systems are phased so the first version arrives early.
4. Run or hand over — monitoring, iteration and support on a retainer, or a full handover with documentation and training. Either way, it is yours.

The stack we work in

Models: OpenAI, Anthropic Claude, and open-weight models where cost, latency or data residency call for them. Platforms: Supabase and Postgres, Next.js and React, n8n, Make and Zapier for orchestration where it fits, Cloudflare and Vercel for hosting, and whatever CRM, phone and document systems you already run. We choose boring, mainstream technology on purpose: your next engineer should recognise everything.

Who this is for

Owner-led and mid-sized businesses — typically 10 to 200 people — that have a real workflow to fix and want one accountable team, not a strategy layer and a build layer that blame each other. If you have a precise specification and an internal team to maintain the result, a straightforward dev shop may be cheaper. If you need a second opinion before committing to anything, start with our consulting service instead.

How pricing works

Audit and design is a fixed-scope engagement. Builds are quoted as fixed scope against the hours and dollars the system removes — no open-ended hourly meter. Running the system after launch is an optional monthly retainer covering monitoring, fixes and iteration. We walk through the numbers on the diagnostic call before anything is committed.