tungsten.dev

Customer case study · Underwriting and due diligence

Put the next deal into the model your team already trusts.

A fund taught Tungsten to use its own model, assumptions, and templates. Six hours of teaching produced an evaluation-ready agent, with every number traceable for analyst review.

Operating baseline

4 analysts
on the team
30–40
deals each month
30%
of time filling the model
6 hours
to teach the agent

What the analysts teach

Analysts teach the fund’s model and conventions.

  • Source documents and Argus exports used in the investment review.
  • The fund’s proprietary model and templates.
  • The assumptions and conventions applied in context.
  • Analyst demonstrations of the real work.

What the agent does

Tungsten builds the next model from the fund’s sources.

  • Populate the proprietary financial model.
  • Apply the fund’s assumptions.
  • Run the required sensitivities.
  • Trace every number to its source.

Agent output

The fund’s workbook, ready for analyst review.

modelApproximately 145K unique formulas remain in the fund’s own workbook.
assumptionsSource references, assumptions, and calculation issues travel with the workbook.
traceEvery populated number links back to its source.

Customer outcomes

Return model-filling time to investment judgment.

01

Model-filling time returned

Reduce the 30% of analyst time spent filling the financial model.

02

Evaluation ready in six hours

Reach an evaluation-ready agent after six hours of teaching.

03

Traceable review

Keep every number traceable to source for analyst review.

Start with one proprietary model

Teach the model-filling work.

Bring an analyst, representative source material, and the model your team already uses.

tungsten.dev