AVISE · A HOUSE FOR YOUR FIRM'S DEAL FLOW
AI IN FINANCE
AI in finance is mostly unglamorous: reading documents, flagging odd transactions, scoring risk, drafting reports. Here is where it is actually used, what the surveys and regulators say, where it fails, and what it means for a small deal team.
Short answer: AI is used in finance to read and summarise documents, detect fraud and money laundering, score credit risk, optimise trading and portfolios, draft reports, and, on the deal side, screen targets and speed up due diligence. Adoption is now the norm rather than the experiment. The open questions are about control, not capability: who checks the output, and what happens when it is wrong.
The Bank of England and FCA survey found that 75% of UK financial firms were already using AI, with a further 10% planning to within three years. The same report found that a third of all AI use cases were third-party implementations, and that foundation models made up 17% of use cases.
CHECKED 30 SEP 2026 · Bank of England and FCA, AI in UK financial services 2024
Inside the finance function itself, Gartner found 59% of finance teams using AI in 2025, against 58% in 2024 and 37% in 2023, in a survey of 183 CFOs and senior finance leaders. Growth has flattened, and data quality and skills are the barriers named. Gartner’s own page blocked automated access, so this is cited from CFO Dive’s report of it.
CHECKED 30 SEP 2026 · CFO Dive reporting Gartner's finance survey
Read those two together. Adoption jumped quickly, then plateaued once the easy wins were taken, and what separates firms now is whether their data and processes are clean enough for the tools to work.
The main uses of AI in finance, what it does in each, and where a person still decides.
USE
Fraud and AML
USE
Credit and underwriting
USE
Trading and portfolios
USE
Reporting and FP&A
USE
Deal screening and diligence
| USE | WHAT THE AI DOES | WHERE A PERSON DECIDES |
|---|---|---|
| Fraud and AML | Flags unusual transactions and patterns across large volumes | Whether a flagged case is real, and whether to file a report |
| Credit and underwriting | Scores applicants, including from cash-flow data | Policy thresholds, exceptions, and explaining a refusal |
| Trading and portfolios | Signals, execution and portfolio optimisation | Risk limits and strategy |
| Reporting and FP&A | Extracts figures from documents, drafts variance commentary | What the numbers mean and what goes to the board |
| Deal screening and diligence | Reads CIMs and data rooms, extracts financials, drafts summaries | Whether to pursue the deal, and the price |
Pattern detection over huge transaction volumes is what machine learning does best, which is why fraud and AML sit among the earliest and most established uses. The Bank of England survey lists them among the highest perceived current benefits of AI in financial services. The catch is false positives: a model that flags everything is a model people stop reading.
The appeal is reaching people a traditional score cannot see.
A FICO pilot using alternative data found that as many as half of applicants who had no score received one of at least 620, a common underwriting threshold. The same briefing notes more than 45 million US consumers lack the history for a credit report or score. The figures are from 2023 or earlier.
CHECKED 30 SEP 2026 · Federal Reserve Bank of Kansas City
A FinRegLab study found that a machine-learning model combining credit bureau and cash-flow data was the most predictive overall and across all subgroups it examined.
CHECKED 30 SEP 2026 · FinRegLab, July 2025
Most of a finance team’s week is moving numbers out of documents and into models. This is the least dramatic use and the one with the clearest payoff, because the task is bounded and the answer can be checked against the page it came from.
FINRA’s 2026 oversight report names summarisation and information extraction from unstructured documents as the leading generative AI use case among member firms.
CHECKED 30 SEP 2026 · FINRA 2026 Annual Regulatory Oversight Report
Deal work is document work: a CIM, a set of financials, a data room. That makes it a natural fit, and the surveys show it.
Deloitte’s survey of 1,000 senior corporate and private equity leaders reported that 86% had integrated generative AI into M&A workflows. By stage, 40% used it for strategy and market assessment, and 35% each for target screening and due diligence. 67% named data security as a leading concern.
CHECKED 30 SEP 2026 · Deloitte survey of generative AI in M&A
Bain’s survey of more than 300 M&A executives found 45% now relying on AI, and describes adoption as having more than doubled during 2025.
CHECKED 30 SEP 2026 · Bain M&A Report 2026
For a large firm with an analyst bench, AI is an accelerator. For a team of one to three with no analyst, it is closer to the difference between reading the CIMs and not reading them. A principal who sees twenty inbound deals a week either triages by gut or triages by numbers, and pulling three years of financials out of a PDF used to be an evening.
The useful question is never whether the AI is smart. It is whether you can see where each number came from.
FINRA defines hallucinations as “instances where the model generates information that is inaccurate or misleading, yet is presented as factual information”, and flags AI agents acting autonomously without human validation and approval as a risk.
CHECKED 30 SEP 2026 · FINRA 2026 Annual Regulatory Oversight Report
In April 2026 the Federal Reserve issued SR 26-2, revised model risk management guidance that supersedes and replaces SR 11-7 from 2011. It emphasises a risk-based approach and is aimed mainly at banks with more than $30 billion in assets. Older articles that cite SR 11-7 as current are out of date.
CHECKED 30 SEP 2026 · Federal Reserve SR 26-2
The SEC charged two investment advisers, Delphia and Global Predictions, for false claims about their use of AI, with penalties of $225,000 and $175,000. Chair Gensler called it “AI washing”. If you say you use AI, you have to be doing it.
CHECKED 30 SEP 2026 · SEC press release 2024-36
Under the EU AI Act, systems used to evaluate the creditworthiness of natural persons or set their credit score are classed as high-risk, with an exception for systems used to detect financial fraud. A Gibson Dunn note on the provisional Digital Omnibus agreement says compliance for stand-alone Annex III systems moves to 2 December 2027; check the final text before relying on the date.
CHECKED 30 SEP 2026 · EU AI Act, Annex III
The Financial Stability Board identifies third-party dependence and concentration, market correlations, cyber risk, and model risk, data quality and governance as the main AI vulnerabilities in finance.
CHECKED 30 SEP 2026 · Financial Stability Board, November 2024
Every one of these points the same way: keep a person accountable, keep the source visible, and do not claim more than the tool does.
Avise is deal flow software for small deal teams and search funds. It reads a CIM and extracts the financials, keeps every deal in one place with a data room and a memo builder, and costs $249 a month for the whole firm, not per seat. It does the first pass so you do the thinking; the figures link back to the pages they came from. See the product page or the pricing, and our comparison with the spreadsheet you already have.
MORE READING
AI CRM
What an AI CRM does for a private equity or search fund deal team, where a person still decides, and six questions to ask any vendor before you sign.
AVISE VS CARTA
Carta's Deal CRM is relationship intelligence and pipeline tracking for private capital, sold demo-only with no published price. Avise is a deal execution system with a data room and a memo builder, priced per firm, in public. Sources and dates on every claim.
AVISE VS MADEMARKET
MadeMarket is a CRM and deal management platform for investment bankers. What it does, what it publishes on price, and how it differs from Avise. Sourced.
AVISE VS HUBSPOT
HubSpot is a general sales CRM priced per seat, built for a sales team working leads and quotes, not a deal team working CIMs. Here is where HubSpot earns its place, where it does not, and what it actually costs to bend it into a deal tracker.
AVISE VS 4DEGREES
4Degrees is a relationship intelligence CRM priced per user, built by ex investors for private capital's network-driven deal teams. Here is where it wins, where Avise wins, and how the price lands at your headcount.
AVISE VS DEALCLOUD
DealCloud is built for firms with an operations team to run it. Avise is built for firms without one. An honest comparison, with sources and dates on every claim.
AVISE VS AFFINITY
Affinity is a relationship CRM priced per user. Avise is a deal system priced per firm. At list price the difference decides itself at about two people.
AVISE VS SPREADSHEETS
The spreadsheet is the real incumbent in small-firm deal flow, and it is genuinely good. Here is exactly where it stops working, and when you should stay on it.
§ · · NEXT
Thirty minutes on a live file of yours, and we will say on the call if this is not a fit. The price is on the pricing page, in full, before you have to ask for it.