AI systems by industry
The plumbing is often similar from one sector to the next. The constraints almost never are. What changes between a hotel group and an NHS trust isn't the technology — it's the regulation, the data sensitivity, and what happens when the system gets something wrong.
Does the industry actually change how an AI system is built?
Partly. The underlying components — document extraction, retrieval, drafting, human review — recur across almost every sector. What changes is the constraint set: which regulator is watching, how sensitive the data is, where the system has to run, and which decisions a human must keep.
That distinction matters when you're buying. An agency that only talks about the technology will build you something that works in a demo and fails an audit. An agency that only talks about compliance will sell you a policy document and no working software.
In practice, the sector determines four things: the deployment model (cloud, private tenancy, on-premise, or fully air-gapped), the evidence you need to produce for a regulator, auditor or board, the escalation design — what the system does when it isn't confident — and the hard limits on what it must never decide by itself.
How the nine sectors differ
Typical first builds, the constraint that shapes them, and how long they take. Timelines are from signed scope to live, and assume one workflow rather than a whole-business rollout.
| Sector | Most common first build | Hardest constraint | Typical deployment | Timeline |
|---|---|---|---|---|
| Legal services | Matter intake, conflict checking and disclosure | Privilege and client confidentiality | Private tenancy, matter-level separation | 6–8 weeks |
| Healthcare & care | Administrative and coordination workflows | The line between admin and clinical judgement | On-premise or air-gapped | 8–12 weeks |
| Logistics & operations | Document processing and exception handling | 24/7 running; failures cost money fast | Cloud, integrated with TMS/WMS | 4–8 weeks |
| Financial services | Client onboarding and document handling | Explainability and an evidenced audit trail | Private tenancy or on-premise | 8–12 weeks |
| Hotels & hospitality | Guest enquiry handling and booking admin | Peaks, seasonality and high staff turnover | Cloud, integrated with the PMS | 4–6 weeks |
| Public sector & charity | Correspondence triage and reporting | Transparency, auditability, procurement | On-premise or air-gapped | 8–12 weeks |
| Property & estates | Enquiry triage and compliance records | Fair treatment of applicants and tenants | Cloud, integrated with the CRM | 4–6 weeks |
| Education & training | Administrative load and content preparation | Children's data and safeguarding | Private tenancy, UK-hosted | 6–10 weeks |
| Retail & e-commerce | Customer messaging and product data | Peak-season volume; brand voice at scale | Cloud, integrated with the storefront | 4–6 weeks |
Which constraint drives your build — data sensitivity, regulation, or volume?
Most projects are shaped by one dominant constraint. Identifying which one you have is the fastest way to know what your build will cost, where it will run, and how long it will take.
Personal, clinical, financial or otherwise confidential data that cannot leave your control. Expect on-premise or air-gapped deployment, a Data Protection Impact Assessment before go-live, and a longer timeline.
Healthcare & care · Legal services · Public sector
A regulator who will ask how a decision was reached. Expect explainability requirements, immutable logging, documented human oversight, and EU AI Act risk classification as a first step.
Financial services · Education · Property & estates
High throughput, seasonal peaks, or work running around the clock. Expect the design effort to go into exception handling and escalation rather than the happy path.
Logistics & operations · Retail & e-commerce · Hospitality
Sector pages, in detail
Each one covers what we'd build, what we won't, and the regulation that shapes it. We're publishing the remaining sectors from the table above over the coming weeks — if yours isn't up yet, just ask us.
What if your industry isn't listed?
It's usually not a problem. The nine sectors above are the ones we're asked about most often, not the limit of what we work on. The questions we'd ask you are the same regardless of sector: what the work is, how sensitive the data is, who is accountable for the output, and what happens when the system is wrong.
If you can answer those four, we can scope a build — whether or not your industry has a page here. Tell us what the work is and we'll tell you honestly whether AI is the right tool for it.
Start with the problem, not the technology.
Tell us what's slow, repetitive, and costing more than it should. We'll tell you whether AI can fix it — and if it can't, we'll say so.
Let's talkLast reviewed: February 2026