Built for one company's work.

Agents, retrieval and the systems around them, for the processes no packaged product fits.

Case study · European Space Agency

It used to take a specialist. Now it takes a question.

SentiBridge is an agentic toolbox we built for the European Space Agency. The answers operators needed were real but buried in thousands of pages of mission documentation, instrument specifications and schemas. They ask in plain language now, and the answer arrives with its source attached.

A Copernicus Sentinel-2 image of South Georgia Island: snow-covered mountain ridges and glaciers running the length of the island, surrounded by dark ocean.
South Georgia Island · Copernicus Sentinel-2 · ESA, CC BY-SA 3.0 IGO. The kind of mission whose documentation SentiBridge answers from.
SENTINEL-CLASS MISSION Mission documentation Instrument specifications Schemas and telemetry Operations logs Agentic toolbox DOCUMENT RETRIEVAL SPECIFICATION LOOKUP UNIT CONVERSION CALCULATION EACH ONE NARROW AND TESTABLE A plain-language answer The source it came from OR "I DON'T KNOW", WHICH IS ALSO AN ANSWER
The shape A toolbox, not a chatbot

The agent answers nothing from memory. It reads the question, picks its tools, runs them, and assembles the answer from what came back. Each tool is narrow enough to test on its own.

The discipline Every answer carries its source

An answer about a spacecraft is worthless if you cannot check it. Each response resolves to the document and section it came from, and says so when the evidence is not there.

The transfer Expertise stops being the bottleneck

The people who knew where things were written down were the constraint on everyone else's work. The toolbox turns that into something the whole team can query.

European Space Agency Sociaw Capgemini

What we build

Twelve jobs, in the words the people doing them would use.

Grouped by sector, though the shapes repeat across all three. Each earns a build only where the volume, the cost per instance and the decision are genuinely there.

Insurance · Intake Take a first notice of loss at 2am

Answered on the first ring, checked against the policy, and in the system of record before the caller hangs up. Anything contested goes straight to a person.

Insurance · Decision support Route a claim to the right handler, with reasons

Severity and specialism scored from the notification and the history, with the evidence for each factor attached. It ships in shadow mode first.

Insurance · Document intelligence Turn a broker submission into structured risk

Schedules, loss runs and covering emails in no fixed format become checked fields, with a confidence per field and a queue for anything below threshold.

Insurance · Review at scale Find the claims you should have recovered

A pass over closed files for missed subrogation, duplicate payment and indemnity leakage, producing a ranked list with the evidence. It does not move money.

Banking · Document intelligence Check a KYC file is actually complete

Every document present, in date and consistent with the others. The output is a checklist naming the discrepancies. It is not an approval.

Banking · Drafting Draft the exception memo

Narrative, financials, covenant position and comparable precedent, assembled into the memo your committee already expects with every figure traceable to source.

Banking · Alert triage Work the alert queue in priority order

Alerts enriched and grouped by likely typology, with a draft narrative and a suggested disposition. The analyst's decision stays the record.

Banking · Change tracking Know which policies a rule change breaks

A regulatory change goes in; the affected policies, procedures and controls come out, mapped and dated. It will not tell you what to do about them.

Legal · Playbook review Mark up a contract to your own positions

Deviations from your playbook, the fallback you usually accept, and the clauses you have no position on yet. That last list is the useful one.

Legal · Review at scale Read the data room and produce the issues list

Thousands of documents into a cited, ranked issues list with the extract that raised each one. Gaps are surfaced rather than smoothed over.

Legal · Retrieval Answer “have we argued this before?”

Precedent and prior positions out of your own matter files, with citations and permissions intact. Often a Company Brain question rather than a build.

Practice · Drafting Draft the scope from work you have already done

Prior engagements, rates and assumptions into a first pass for a partner to correct. The least glamorous use on this page, and one of the fastest to pay for itself.

Agents

An agent is a model that can act, and stop.

Not a chatbot with a better prompt. It reads a request, picks tools, runs them, checks what came back, then answers or hands over. Real capability, and the most over-prescribed architecture in the market.

Anatomy Tools, not knowledge

The power is the toolbox: narrow, individually testable functions. SentiBridge is that shape. Broad clever agents are hard to debug; narrow tools behind a planner are not.

Skills Triage, respond, confirm

The market has settled on a few composable skills, mixed per use case. Useful because each can be evaluated on its own, and because confirm is where the risk gets contained.

Control A determinism dial

How much latitude the agent gets should be a setting, not a personality. Generous where phrasing matters, hard-coded where policy does, decided per workflow.

Escalation Handoff with the summary written

Knowing when to stop is a feature. It routes to the right person and hands over what it already established, so the human starts from the middle of the problem.

Where they earn it Front office, back office, engineering

Lead enrichment and follow-up in sales. First line support across chat, phone and WhatsApp. Onboarding, reconciliation and exceptions in the back office. Specification questions in engineering.

Retrieval

Talking to your documents is the easy part.

RAG in one line: find the relevant passages, put them in front of the model, make it answer from those rather than from memory. A demo takes an afternoon. Three decisions separate that demo from something a business can rely on.

Ingestion Layout-aware chunking

Split so a clause or a specification table is never cut in half. In document-heavy work this decides more of your accuracy than the choice of model does.

Ranking Cast wide, then narrow hard

Fifty candidates from hybrid keyword and vector search, re-ranked down to five. Re-ranking is usually the step where a naive pipeline starts working.

Second passes Search again rather than guess

If the evidence comes back thin the system searches again instead of answering anyway. The difference shows up in the cases someone will later ask about.

Boundary Per project, or shared

Retrieval built for one use case is fine until the fourth, at which point you maintain four disagreeing versions of the truth. Then it should become a layer everything reads from.

How you know it works

Five questions, five different layers.

Whichever job you pick, this is the part that decides whether you can defend it. The layer teams skip is almost always the one that fails silently.

Evals "Is this good enough to ship?" Golden corpus · LLM-as-judge · groundedness · CI gates
Observability "What actually happened in production?" OpenTelemetry traces · prompt versioning · cost per call
Guardrails "Is this response safe to send, right now?" Inline validators · injection containment · refusal policy
Rollout "Can we ship without betting the whole queue on it?" Per-cohort A/B · auto-rollback · gateway budgets · P99, not the mean
Portability "Can we leave?" OpenTelemetry-native · self-hostable · open standards over lock-in