Thesara AI / Services
Assessment, custom development, migration off legacy systems, integration with what you already run, and the operating support that keeps it working after go-live. Fixed scope, measured results, no open-ended discovery.
The state of play
The gap between an impressive demonstration and a production system is not model quality. It is everything around the model — permissions, data access, exception handling, evaluation, audit, and the unglamorous work of changing how a team actually operates.
That gap is why Gartner expects more than 40% of agentic AI projects to be cancelled by 2027. Almost none of those cancellations will be because the model could not do the task. They will be because nobody could prove who did what, because the pilot was measured against a moving target, or because the system was built for a demonstration and could not survive contact with real data.
Our services practice exists to close that gap. We have built the trust layer, the evaluation harness and the audit trail once, for our own products — so a client engagement starts with those already solved rather than discovering them in month four.
Service lines
Most clients start with the first and progress through the rest. Each one has a defined deliverable and a defined end.
A fixed-fee review of your systems, data, workflows and compliance obligations, ending in a ranked list of candidate workflows with an estimated value and difficulty for each.
Typical duration: 3–5 weeks · Fixed fee
Building the application itself, on the Thesara platform, for workflows that no off-the-shelf product covers. You own the resulting application and its data.
Typical duration: 3–9 months · Fixed scope per phase
Moving a workflow off a client–server or first-generation cloud system without a cutover weekend that everyone dreads. The old system keeps running until the numbers match.
Typical duration: 4–12 months · Phase-gated
Connecting AI capability to the systems you are keeping, and training your own team to extend it. The goal is that you need us less over time, not more.
Typical duration: 6–16 weeks
An AI application is not finished at go-live. Models change, data drifts, and a system that was accurate in March can be quietly wrong by September. This retainer covers continuous evaluation, incident response, model updates and the periodic evidence pack your auditors will ask for.
Ongoing · Monthly retainer
Engagement model
Every engagement follows the same shape, whatever the industry. The success measure is agreed in step two, in writing, before any building begins.
We map the workflow as it actually runs — not as the process document describes it — and establish a baseline: how long it takes now, how often it goes wrong now, what it costs now.
One workflow, one team, one quarter, and a specific measurable target both sides sign. This is the step most vendors skip, and skipping it is why most pilots end in an argument about whether they worked.
Delivered in two-week increments against the Thesara platform, with the trust layer configured first. You see working software from the third week, not a slide deck at the end.
Run against the baseline with real work and real users. The result is reported as it is. If the target is missed, we say so plainly and set out what would change the outcome — or recommend stopping.
If the number was met, the choice is to extend into adjacent workflows or move the system onto a managed operation retainer. If it was not, the engagement ends with a written account of why.
Delivery principles
Identity, permissions, residency and logging are configured before the first feature. It is the only order that reaches production.
Every engagement ships a test set and a harness the client keeps. Without it, nobody can tell whether next quarter's model update helped or hurt.
Dosage limits, credit thresholds, price floors and safety rules are code, not prompts. A model may propose; it may not override.
We design what happens when the system is unsure before we design the happy path. That queue is where the humans do their real work.
Client data is not used to train models for anyone else. Ever. It is written into the contract, not just the policy page.
Documentation, handover and developer enablement are in scope by default. A client who cannot maintain what we built is a failed engagement.
Commercials
| Service line | Pricing shape | Commercial role |
|---|---|---|
| AI Readiness Assessment | Fixed fee | Low margin by design — it is how an account opens |
| Custom Development | Fixed scope per phase, milestone-billed | Core project revenue; predictable and staffed |
| Legacy Migration | Phase-gated fixed price | Largest engagements; longest client commitment |
| Integration & Enablement | Fixed fee or capped time-and-materials | Shortens the path to a subscription |
| Managed Operation | Monthly retainer | Recurring, high margin, grows without headcount |
| Suite subscription | Annual per seat, site or outcome | The destination — where services revenue is meant to lead |
Talk to us
Tell us the process that costs your organisation the most time and goes wrong the most often. That is usually the right place to begin.