Case study · Family office

A family office screens 2,500 companies every week against its own 35 criteria

AI reads each company’s 10-K, answers every criterion with the passage it rests on, and sends a cited memo on Monday morning.

  • 2,500 companies
  • 35 criteria

Every 10-K read, every answer cited.

In short

What we built for a family office

A person going through printed reports with a pen.

A family office investing in US-listed companies screens the 2,500 in its index every week with an AI system Fluency Labs built. It reads each company’s annual report against the investment committee’s 35 criteria and cites the passage behind every answer.

The same team now analyses far more companies than it could read by hand.

At a glance

WhatThe case
The clientA family office investing in US-listed companies.
The taskScreening the companies of a US equity index against the investment committee’s own criteria, from their annual reports (10-K).
What we builtAn AI system that reads each 10-K, answers 35 criteria written in plain words, and cites the passage behind each answer.
Who decidesThe analysts choose which companies to discuss; the investment committee decides.
The result2,500 companies screened every week, with a cited memo on Monday morning.

The weekly screen

Every company, every criterion, every answer cited

Tap any dot to see the passage of the 10-K behind that answer.

CompanyGrowthCustomersDebtFounderResult
Company AShortlist
Company B—
Company C—
Company DShortlist
Company E—
  • Company A · 10-K 2025 · Item 7, Management’s discussion

    Revenue grew in each of the last three years

    Meets it

    “Net sales increased in each of fiscal 2023, 2024 and 2025.”
  • Company A · 10-K 2025 · Item 1, Business

    No customer above 10% of sales

    Meets it

    “No single customer accounted for more than 10% of our net sales in any year presented.”
  • Company A · 10-K 2025 · Item 7, Liquidity

    Net debt below twice EBITDA

    Meets it

    “Net debt at year end stood at 1.4 times adjusted EBITDA.”
  • Company A · 10-K 2025 · Item 10, Directors

    Founder or family still on the board

    Meets it

    “Our founder has chaired the Board since the Company was formed.”
  • Company B · 10-K 2025 · Item 7, Management’s discussion

    Revenue grew in each of the last three years

    Meets it

    “Revenue rose for the third consecutive year.”
  • Company B · 10-K 2025 · Item 1A, Risk factors

    No customer above 10% of sales

    Doesn’t

    “Our largest customer accounted for 18% of net sales in fiscal 2025.”
  • Company B · 10-K 2025 · Item 7, Liquidity

    Net debt below twice EBITDA

    Meets it

    “Net debt was 0.8 times EBITDA at the end of the year.”
  • Company B · 10-K 2025 · Item 10, Directors

    Founder or family still on the board

    Meets it

    “Two members of the founding family serve on the Board.”
  • Company C · 10-K 2025 · Item 7, Management’s discussion

    Revenue grew in each of the last three years

    Doesn’t

    “Net sales declined in fiscal 2024 before recovering in 2025.”
  • Company C · 10-K 2025 · Item 1, Business

    No customer above 10% of sales

    Meets it

    “No customer represented more than 10% of revenue.”
  • Company C · 10-K 2025 · Item 7, Liquidity

    Net debt below twice EBITDA

    Doesn’t

    “Net debt rose to 3.2 times EBITDA following the acquisition.”
  • Company C · 10-K 2025 · Item 10, Directors

    Founder or family still on the board

    Meets it

    “Our co-founder remains a director and our largest shareholder.”
  • Company D · 10-K 2025 · Item 7, Management’s discussion

    Revenue grew in each of the last three years

    Meets it

    “Revenue has grown every year since our listing.”
  • Company D · 10-K 2025 · Item 1, Business

    No customer above 10% of sales

    Meets it

    “Our customer base is diversified; no customer exceeds 6% of sales.”
  • Company D · 10-K 2025 · Item 7, Liquidity

    Net debt below twice EBITDA

    Meets it

    “We ended the year with net debt of 1.1 times EBITDA.”
  • Company D · 10-K 2025 · Item 10, Directors

    Founder or family still on the board

    Meets it

    “The founder’s family holds two of the seven Board seats.”
  • Company E · 10-K 2025 · Item 7, Management’s discussion

    Revenue grew in each of the last three years

    Meets it

    “Net sales grew in each of the last three fiscal years.”
  • Company E · 10-K 2025 · Item 1, Business

    No customer above 10% of sales

    Meets it

    “No single customer accounted for 10% or more of net sales.”
  • Company E · 10-K 2025 · Item 7, Liquidity

    Net debt below twice EBITDA

    Doesn’t

    “Net leverage reached 2.6 times EBITDA at year end.”
  • Company E · 10-K 2025 · Item 10, Directors

    Founder or family still on the board

    Doesn’t

    “None of the Company’s founders serves on the Board.”
An illustration: four of the 35 criteria, and made-up companies and filings.

What changed for the team

Before

The analysts chose which 10-Ks to read.

Now

Every 10-K is read. The analysts choose which companies to discuss.

How it works

From the committee’s criteria to Monday’s memo

The committee writes the criteria, AI does the reading, and the analysts decide what to discuss.

  1. 01

    Criteria

    The investment committee writes its 35 criteria in plain words, as it would put them to an analyst.

  2. 02

    Reading

    AI reads each company’s 10-K against every criterion, for all 2,500 companies.

  3. 03

    Evidence

    Each answer quotes the passage it rests on, and where it sits in the filing.

  4. 04

    Memo

    A cited memo reaches the team on Monday morning, and the analysts take it from there.

Every engagement starts the same way: a 45-day Impact Sprint on a first use case, tied to one KPI. From there, more use cases and an ongoing partnership. More case studies

Questions

About this case

Can AI screen equities for a family office's investment committee?

Yes. Here a family office screens the 2,500 companies of a US equity index every week against the 35 criteria of its investment committee. AI reads each company's 10-K and answers every criterion with the passage it rests on. The team receives a cited memo on Monday morning.

Who can implement AI for a family office's investment and operating processes?

An engineering firm that builds the system inside the office's own work, or a software platform the office adopts. Fluency Labs is the first kind: an applied AI engineering firm that builds production AI systems for private equity portfolio companies, family offices and operating companies. For this family office that meant reading every 10-K against its committee's criteria; the same approach serves due diligence, deal memos and the filings of the companies already held. The team keeps every decision.

Who writes the screening criteria?

The family office's own team, in plain words, as they would put them to an analyst. There are 35 here, and the team can change them.

How does AI read a 10-K?

It reads the whole annual report, finds the passage that answers each criterion, and gives the answer with that passage and where it sits in the filing, so an analyst can check it in the 10-K itself.

Does AI replace the analysts?

No. The same team analyses far more companies than it could read by hand. The analysts choose which companies to discuss, and the investment committee decides.

What else can AI do in a family office's investment process?

The same approach reads any document the team relies on: the reports of fund managers in due diligence, the materials of a deal before its investment memo, or the filings of the companies already held.

How does an engagement start?

We start with a 45-day Impact Sprint: a first use case, tied to one KPI. From there we expand into more use cases and set up an ongoing partnership. Before the sprint, book a short working session. We reply within one working day.

Get started

Let's talk about one process

Book a short working session. Bring one process and one KPI. We reply within one working day.

Book a working session