Four problems we are asked to take on more often than any others. Each one sets out the failure mode, the order we would work it in, what is live at the end of the first fortnight, and what you would own when we leave. They are starting points rather than packages, and most engagements diverge from them in week one once we have seen your systems. That is the point of starting from a brief rather than a blank page.
The engagement patterns we are asked for most, across AI, web, mobile, cloud and security.
A support agent grounded in your own documentation, with a confidence floor that escalates to a person rather than guessing. Built from the tickets you already have, so the intents it handles are the ones that genuinely repeat, and shadow-run against live traffic before it answers anybody.
Old and new running side by side until the new path has earned the traffic. Nothing is switched on a single weekend: routes move a slice at a time, differences are investigated rather than waved through, and the previous path stays available the whole way through.
A platform where access control, lineage and audit are part of the design rather than bolted on before an inspection. Every field traceable to its source, every permission enforced at the row and column, and the evidence an auditor asks for produced without a scramble.
A copilot over internal knowledge that respects the permissions people already have, so it can never surface a document someone was not allowed to open. Answers cite their sources, staleness is visible, and the questions it cannot answer well are the ones it hands to a person.
Contracts, statements and forms turned into structured data you can query, with a confidence score on every field and a human in the loop wherever it is low.
Search and discovery rebuilt around what customers actually type, blending keyword and semantic matching, with relevance you can measure rather than argue about.
The repetitive internal process mapped, automated where it is safe, and left with a person wherever judgement or liability sits.
A fixed-price assessment of whether the AI work you are considering will actually function on the data and systems you have, before you spend a quarter finding out.
A slow, dated self-service portal replaced incrementally, with the old one still serving customers until each new page has earned its traffic.
A mobile app for people working where the signal is not, with capture that keeps working offline and sync rules that resolve conflicts predictably.
An inherited cloud estate mapped, attributed and right-sized, with the guardrails that stop the bill drifting back up once attention moves on.
Security moved into the pipeline rather than into a review at the end, with scanning, secrets management and rollback that have actually been rehearsed.
None of these will match your situation exactly, and they are not meant to. They are the briefs we are asked for most often, which makes them a faster place to start an argument than a blank page.
| Solution | Usually asked for by | To first release | Choose this when |
|---|---|---|---|
| Agentic Support Deflection | SaaS & Commerce | 2 weeks | Inbound support volume is growing with the customer base and a handful of questions account for most of it. |
| Legacy Modernisation, No Downtime | Financial Services | 2 weeks | A system runs the business, the people who wrote it have gone, and no one will sign off a big-bang cutover. |
| Data Platform Under Regulation | Healthcare & Public Sector | 2 weeks | An auditor, a regulator or an enterprise buyer is about to ask questions your current platform cannot answer. |
| Internal Copilot on Private Knowledge | Enterprise & Operations | 2 weeks | Your people spend real time hunting for answers that already exist somewhere internal, behind permissions that must be respected. |
| Document Intelligence Pipeline | Financial Services & Legal | 2 weeks | People are rekeying the same document types by hand, and the errors only surface downstream. |
| Search That Understands Intent | E-commerce & Marketplaces | 2 weeks | Your search returns nothing useful for the terms customers really use, and nobody can prove whether a change helped. |
| Back-Office Workflow Automation | Operations & Shared Services | 2 weeks | A process runs on spreadsheets, email and one person who knows how it really works. |
| AI Readiness and Data Audit | Any sector, before you commit | 2 weeks | A board or a customer is asking for AI and nobody internally can say honestly whether it is feasible yet. |
| Customer Portal, Rebuilt | B2B & Services | 2 weeks | Customers avoid your portal and call your support team instead, and a full rebuild feels too risky to start. |
| Offline-First Field App | Field Operations & Logistics | 2 weeks | Your field team records work on paper or in a spreadsheet because the app stops working the moment they lose signal. |
| Cloud Cost and Reliability Reset | Any sector on public cloud | 2 weeks | The cloud bill grows faster than usage and nobody can explain which team or feature is responsible. |
| Secure Delivery Pipeline | Regulated & Enterprise | 2 weeks | Security review is a gate at the end of delivery that everyone dreads, and releases slow to a crawl because of it. |
Agreed before anything starts, so the number you approve is the number you pay. Changes are a conversation, not an invoice you find out about later.
Your repository, your cloud tenancy, your licences. There is no runtime of ours you have to keep paying for, and nothing to migrate when we are done.
Deployed and usable, not demonstrated in a meeting. You can stop at the end of any cycle and keep everything built up to that point.
Agreed at the start, whether that is answer quality, latency, cost per request or reconciliation accuracy. You can re-run it yourself after we leave.
A solution is a starting point, not a fixed package. The shape of the engagement is the same either way, and you can stop at the end of any fortnight.
A mutual NDA if you want one, then a call to work out whether the closest solution is actually close enough. Sometimes the honest answer is that your problem needs its own scope, and we will say so.
We read what exists: the systems, the data and the constraints nobody wrote down. The plan gets adjusted to what is actually there rather than to what the brief assumed.
The riskiest part is built first. If retrieval will not be accurate enough, or an integration will not hold, you find out in week one rather than in month three.
A release every two weeks against the measure agreed at the start, with your team working alongside ours so the handover is continuous rather than an event at the end.
What people ask when they are deciding whether to start from one of these or from a blank page.
No. They are starting points, not packages. Most engagements begin with one of these and diverge in the first week once we have seen your systems, which is the point of looking at them first rather than starting from nothing.