Two years of AI pilots in financial services. The reporting cycle is still measured in weeks.
DDQs still answered from memory and old files. Relationship managers still spending client hours on assembly work. The pilots were real and the tools are live; what most of them shipped without is the verification and adoption layer that turns output a compliance officer would challenge into output the institution can act on. Building that layer — into the tools you have, or the ones you should — is our work.
Financial institutions have spent the past two years piloting AI, and most of the attention has gone to the investment and risk functions. The faster payback is usually elsewhere: in distribution, reporting, and client service, where senior professionals spend a striking share of their time assembling documents rather than exercising the judgement they were hired for. Investor-relations teams still describe diligence response as the most operationally punishing part of fundraising — despite a market with more than thirty tools selling into exactly that workflow.
That abundance moves the COO's question from whether the technology works to which workflow to engineer first, with which tool — bought or built — under which data and governance constraints, adopted by a team whose every output may be read by a client, a regulator, or both. Judgement questions of that kind are poorly served by anyone with a product to sell, and several of the products have usually been bought already.
We are based in Basel, work in the regulatory environments our clients answer to, and treat governance as an architecture decision made early — not a compliance document written after the fact.
Four places AI pays back on the client-facing side of a financial institution.
RFP and DDQ response
Institutional questionnaires drafted from your approved record — prior submissions, compliance language, current data — for your experts to review rather than write. The team that wins mandates stops losing its weeks to the paperwork that accompanies them.
Client and investor reporting
Reporting cycles compressed from weeks to days, with commentary drafted against the numbers rather than assembled around them. Few asset managers believe their reporting meets client expectations; most agree it wins or loses mandates. Closing that gap is now engineering work rather than a staffing decision.
Relationship-manager capacity
Meeting preparation, portfolio review assembly, suitability documentation — the hours of gathering that precede every client conversation, done before the RM sits down. The client gets more of the person they chose the firm for; the firm gets more client conversations from the same headcount.
Onboarding and file assembly
KYC packs, credit files, claims documentation: the document-heavy assembly work where institutions lose days and applicants lose patience. AI extracts, validates, and flags inconsistencies before a human reviews the file — review by exception, not by default.
The structural problem behind a DDQ is not unique to finance: a regulated supplier answering institutional questionnaires from years of prior submissions, where the language must be exact and the expert should review rather than write. We engineered precisely that motion for a €160M life-sciences supplier responding to pharmaceutical procurement tenders — four years of submissions, quality agreements, and regulatory filings indexed into a knowledge base an AI agent drafts from.
In wealth itself: we designed the document-intelligence architecture behind a cross-border wealth platform — six asset-document classes parsed into a unified portfolio view, with PII and financial data separated at the storage layer so the AI operates identity-blind by design, not by policy. Privacy architecture before any code.
Every engagement runs on the same arc — diagnose, harness, calibrate, demonstrate. We start with one piece of real work: a DDQ, a reporting cycle, an onboarding file, traced end to end, asking where a right answer comes from and what a wrong one costs. That diagnosis takes days and tells you whether the investment is recoverable and where the value actually sits — usually in the response, reporting, and preparation hours that go to assembly rather than judgement.
Where we build, the data architecture comes first — what the AI may read, what it may never see, and how that separation is enforced structurally rather than by policy — and the output ships inside a harness: a drafted response checked against your approved record and the regulatory language that must be exact, so a plausible answer and a correct one stop being indistinguishable. We are independent of every vendor in this market, including the thirty selling DDQ automation, and we stay embedded through the first live cycles until your team trusts the output without us in the room.
The partner who scopes the engagement runs it.
What COOs and CDOs in financial services usually ask first.
We've piloted AI for two years. Why is the reporting cycle still measured in weeks?
Because most pilots shipped without the verification and adoption layer that turns output a compliance officer would challenge into output the institution can act on. The pilots were real and the tools are live — the gap is between a plausible draft and one your team will actually put its name on.
With thirty-plus vendors selling DDQ and RFP automation, how do we choose?
That abundance moves the question from whether the technology works to which workflow to engineer first, under which data and governance constraints — a judgement question poorly served by anyone with a product to sell, especially since several of the products have usually been bought already. We're independent of every vendor in this market, including the thirty selling into this exact workflow.
How do you handle data privacy and governance in a regulated environment?
Governance is an architecture decision made early, not a compliance document written after the fact. We specify what the AI may read and what it may never see before any intelligence layer is built, and enforce that separation structurally — in one build, PII and financial data are separated at the storage layer so the AI operates identity-blind by design, not by policy.
What return has this produced for a comparable firm?
For a €160M life-sciences supplier responding to pharmaceutical procurement tenders — the same structural problem as a financial-services DDQ — four years of submissions, quality agreements, and regulatory filings were indexed into a knowledge base an AI agent drafts from. Tender response time fell from three weeks to four days, tenders per quarter rose 70%, and expert hours per bid fell 65%.
Bring us the questionnaire that ate last quarter, or the report that ships late every month.
Twenty-five minutes. We'll tell you whether the pattern transfers to your shop, and we'll be just as clear if it doesn't.
Schedule a call