Applied AI for Finance

Prove your finance team can safely apply AI to real work — in 30 days.

A facilitator-led training program for your finance team: A baseline task, live practice on real finance cases, graded submissions, and a management report on what the team can do.

View training scopeSee a sample capability report
Participant rating across two pilot cohorts, in a global reinsurer's finance function
4.8/5
Participant rating across two pilot cohorts, in a global reinsurer's finance function
Very or extremely confident using AI, before vs. after
25% → 69%Very or extremely confident using AI, before vs. after
CAPABILITY REPORT · YOUR TEAMSAMPLE

Before and after

58 → 86

median score, baseline vs. final task

4 → 19

of 24 people at a level a reviewer accepts

AI capabilities

BeforeAfter
Accuracy & facts70% → 92%
Format & house style64% → 88%
Materiality judgment49% → 74%
Management actions12% → 61%
ILLUSTRATIVE DATA

30-day Applied AI training scope

Measure, practice, measure again

Your team starts with a scored task, practices on real cases in facilitated sessions, and finishes with a comparable task. The report shows what changed.

Baseline task: A scored task before the first session

Build, break, harden: Live sessions on your cases

Graded submissions: Scored by code, live on screen

Final task: Same type of task as baseline

Capability report: Before/after scores and next steps

APPLIED AI TRAINING SCOPE
30
DAYS
2 × 4h
LIVE SESSIONS
15–35
PARTICIPANTS
Fixed
FEE PER COHORT
Participant effort
The two live sessions, plus the baseline and final tasks. Casework takes place during the sessions.
Cases
Selected with your team from the use-case library. Additional cases can be developed with you for your program.
Tools & data
Your approved AI tools, our synthetic data. No sensitive company data is required at any point.
Deliverables
The capability report, per-participant graded feedback, the baseline-to-final comparison.
Pricing
Quoted at the team-fit call and confirmed in writing before you commit.
Guarantee
If the before-and-after tasks show no measurable improvement, we run another session cycle free.

Learn, then build it

Step 1: Build the capabilities in the applied AI training

A month of baseline measurement, live casework, and graded practice to learn how to apply AI to your team’s work.

Check the fit

Step 2: Apply to real use-cases in the accelerator

Bring the use case that is most relevant to you. We build it with your team into a working MVP that belongs to you.

Discuss the accelerator

A self-serve subscription is in development & pilot clients get early access. Join the waitlist.

Cases & method

The use case library

The applied AI training runs on the use cases your team picks.

UC-04Blueprint

Check Failure Triage

Groups quarterly control-check failures by cause, drafts a proposal, and routes tickets to owners.

Classification & triage
UC-05Blueprint

IGR Contract-to-Cashflow

Reads retrocession contracts and extracts parameters with clause-level citations.

Extraction
UC-06Blueprint

Closing Calendar Orchestrator

Tracks closing dependencies, names what's blocked downstream, and drafts the follow-up emails.

Orchestration
UC-07Blueprint

Ticket Classification

Classifies messy free-text tickets while code calculates volumes and wait times.

Classification & triage
UC-08Blueprint

Semantic Layer Readiness

Checks that every metric has one governed definition: aliases, sign conventions, source priority.

Foundations
UC-09Blueprint

Verification Playbook

How a finance team verifies AI output without redoing the work by hand: deterministic gates and evals against known answers.

Controls & verification
UC-10Blueprint

IFRS 17 Agentic Tutor

A tutor grounded in a calculation engine that makes juniors predict outcomes before explaining them.

Expertise
UC-01 · RECONCILIATION & CONSISTENCY CHECKER

THE TASK

The investor slide reports group net premium earned of USD 11.2bn. Does it agree with the governed source figure before the deck goes public?

How it runs today

An analyst reads the slide against the disclosure pack, recomputes the rounding by hand, and a second person repeats the same check before sign-off.

Most quarters everything agrees, and both people have spent the afternoon proving it.

With the agent
Match

Net premium earned extracted from the slide SLIDE p.4 and retrieved from the governed source FPSL Q1.

Code compares 11.2bn against 11.184bn within the agreed rounding tolerance and writes the report line with both citations. No judgment call was delegated to the model.

RESULT

12 KPIs verified

The examples above are shortened. Ask for the live demo here

One method behind every case

Every case above follows the same pattern. After the course you can apply it to any use case.

  1. Governed inputs

    You provide the documents, data, and decision rules the agent can use.

  2. Agentic step

    The model handles the interpretation: reading, classifying, extracting, and drafting.

  3. Deterministic checks

    Code recomputes the facts, checks every required output, and blocks mismatches before they reach a reviewer.

  4. Human review

    A person approves, edits, or rejects. Ambiguity is highlighted, never resolved silently.

Evidence

How your team is measured

Progress is scored for every use case across different categories. Below is one example from a real training session: the quarterly commentary exercise, with the live dashboard that participants were graded on and the report it feeds into.

UC-02 · QUARTERLY COMMENTARY · LIVE SESSION 2026-07-07UPDATED 0s AGO
322
SUBMISSIONS
100
TOP SCORE
240
ELIGIBLE
LIVE RANKING
SCORE SPLIT · Erin · 94/100
Driver accuracy24/24
Management actions14/20
Total variance16/16
Format12/12
Style-guide language12/12
Materiality8/8
Regional amounts4/4
Commentary quality4/4

The description is flawless. The 6 missing points are all in “what should management do about it”, exactly the gap the next session works on.

Real session data with anonymized names. The six entries are sampled from real scores. A similar dashboard runs on the classroom screen during every training.

What lands on your desk after 30 days

Before/after performance, capability by area, and what to do next. This sample uses illustrative data. Your report carries your cohort’s measured numbers.

CAPABILITY REPORT · FINANCE TEAMSAMPLE · ILLUSTRATIVE DATA

Before and after

58 → 86

median rubric score on the task, baseline vs. after the program

4 → 19

of 24 participants scoring above 70, the level a reviewer would accept

Illustrative figures showing the report format. Same type of task, scored against the same categories, four weeks apart. Individual before/after scores are in the role-level appendix.

AI capabilities

BeforeAfter
Accuracy & facts70% → 92%
Format & house style64% → 88%
Materiality judgment49% → 74%
Management actions12% → 61%

Aggregate rubric scores across all post-program submissions. The team describes movements accurately. Recommending what to do about them is where practice should continue.

What to do next

  • 01Move from description to decision. Management actions scored 61%. Practice turning a correct description of a movement into a recommendation the team can act on.
  • 02Two promising use-cases identified during the training Check-failure triage and IGR contract extraction came out of the sessions as blueprints. Both are accelerator candidates.
  • 03Name an owner for the next use case. Choose one case to keep moving after the applied AI training. Reuse the same task family and rubric at 90 days to see whether the practice stuck.
ILLUSTRATIVE DATA

Fit & next step

Who the training is for, and who it isn’t

The applied AI training works best when the team had exposure to AI tools and is starting to think about solving problems with AI.

RUN THE APPLIED AI TRAINING IF
  • A leader (CFO, head of finance, controller) wants the capability question answered and will read the report.
  • Participants have a company-approved AI tool, or can get one before day one.
  • The team will complete graded work: Baseline task, live casework, final task.
NOT YET, IF
  • You are looking for an inspiration session with passive attendance (no hands-on use-cases).
  • You expect production systems in 30 days. That’s the accelerator, the step after the applied AI training.
  • There is no approved AI tool in your company. We can advise during the call.
The next step is a 20-minute team-fit call. We check objectives, participant readiness, tool constraints, and success criteria, and quote the fixed fee. If it’s not a fit, we say so.

Request a 20-minute call

Tell us about your team and we'll come back to schedule the call and answer your questions on format, fit, and price.

Prefer email? Write to