Financial services
Evidence of how AI is changing engineering work, built for regulated environments.
Banks and insurers have invested in AI coding assistants and agents. Usage dashboards show the spend. Ascenda shows what is happening to the engineering work and the people directing it, in aggregates of ten or more, with no source code leaving a machine.
Three questions we hear
The same investment, three different questions about it.
A technology leader, an enterprise architect and a transformation lead each arrive with a different question. The answers below are what Ascenda can stand behind today.
The CTO
“We have invested heavily in AI coding assistants and agents. Are we getting what we expected, and what evidence can I take to the executive committee?”
A baseline of how engineering work is composed across your cohorts, from the tools' own records, and a readout of each change you declare, reported whichever way it goes. Evidence about the work, rather than a usage figure or a survey.
Enterprise architecture
“Does this introduce another platform into our environment? What does it collect, and does it need access to source code?”
Nothing to integrate. No repository access, no single sign-on, no CI hook. Engineers install Flow or the open-source collectors on their own machines, and only counts, timings and tool events leave, over outbound HTTPS. The collectors are Apache-2.0, signed, and can be inspected before anything is sent.
AI transformation
“How do we tell genuine change from more activity, and benchmark our progress over time?”
Ascenda separates exposure (how much of the work runs through agents) from composition (what it asks of the engineer) and from what engineers report. A declared change is scored against a frozen baseline, so the readout cannot be fitted to the result. Cross-organisation comparison is being built from organisations that choose to contribute.
What a regulated buyer assesses
Four areas, each with its evidence.
Supplier assessment in financial services runs on four questions: is the AI safe and responsible, is the data handling compliant, is the commercial offer clear, and can it be implemented. Each links to the page that answers it.
Safe and responsible AI
Ascenda observes measurable patterns of work and scores nobody. How the work felt is reported only as what people told us in their own check-ins, in aggregate. Measures are in calibration and we say so.
The published rules →Privacy and compliance
A data inventory, processing boundaries, an Australian hosting region, retention and deletion, and the limits we do not yet claim, set out for a privacy impact assessment. Per-person consent, listed in plain words before anyone joins.
Enterprise readiness →Procurement and commercial engagement
A defined 30-day Report with a written deliverable, a cohort of ten or more, and one setup call. Priced per engagement, with the 90-day Study as the next stage. No subscription is required to start.
The engagement →Implementation
Four steps: setup call, engineers join, thirty days, readout. Installed per user without administrator rights, outbound HTTPS only, and a local-file mode so your security team can read the payload first.
Running a limited pilot →The engagement
Start with one cohort and a 30-day Report.
Pick an engineering cohort of ten or more. After a setup call, engineers who choose to join install Flow or the collectors. The Report can draw on history their tools already hold, so the first read covers weeks already worked. Thirty days later you receive a written report of the cohort’s AI exposure and work composition, baseline findings and a proposed measurement plan.
The 90-day Study is the next stage: live, with engineers’ check-ins in aggregate, and a declared change scored against the frozen baseline. Organisations that take part can help build the comparison across organisations, from people who choose to contribute.
What Ascenda will not do
- Score the productivity or performance of a person or a team.
- Show any individual to a manager, at any level of detail or on request.
- Claim a sector comparison before enough organisations have contributed to one.
- Claim a certification, audit or compliance status we do not hold.
- Send prompts, source code, model outputs or file paths off an engineer's machine.
One question to start with
How does your organisation currently determine whether AI adoption is improving the way engineering teams work?
If the answer today is licence counts and a developer survey, a 30-day Report is the shortest route to a better one.