# AI in M&A: Building, Buying, and the Future of Dealmaking

Kison Patel, DealRoom | MLOps Podcast | Episode 315 | 55:33
Hosted by Demetrios Brinkmann

Source: https://www.youtube.com/watch?v=1QzijeBH7TU
Channel: MLOps Community, now AAIF Live (https://www.youtube.com/@AAIFLive-x1r). Summarised by MLOps Talks.
Page: https://mlopstalks.com/talks/ai-in-m-a-building-buying-and-the-future-of-dealmaking
Published: 2025-05-16
Tags: build-vs-buy, finance, product-strategy

## TL;DR
- Kison Patel found DealRoom's strongest AI use case by speaking with customers and learning that contract analysis could reduce legal expenses.
- DealRoom created a small team around an existing product to give its AI work fewer competing priorities while retaining the larger platform's security, customers, and reputation.
- Kison Patel expects AI company valuations to correct because many products have weak defensibility, high venture-backed exit expectations, and limited evidence of acquisition demand.

## Summary
Kison Patel describes how his M&A background led to DealRoom, after an earlier software business failed when Google treated its linking approach as link farming. DealRoom began as a project management tool for M&A and now supports the buy-side process from pipeline management through diligence, integration planning, execution, and reporting. Patel explains how customer conversations redirected the company's AI work toward extracting information from contracts, especially change-of-control, assignment, and consent provisions. He is candid about the difficulty of adding innovation inside a 50-person company, where AI work competes with customer and sales requests. DealRoom responded by putting the AI roadmap into a small, separately managed product team. The conversation also covers AI pricing, hiring, distribution, and M&A strategy. Patel argues that buyers get better outcomes when acquisitions follow a defined company and M&A strategy. He expects strong SaaS businesses to remain attractive while many AI startups face skepticism and valuation corrections.

## Key ideas
### DealRoom grew from a failed content business and a project-management idea for M&A
[16:57](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1017s)
Kison Patel began in M&A advisory work, first handling buyers and sellers of small businesses, then building his own practice. After moving into technology, he created hundreds of finance-focused domains and automated blogs, but Google later classified the linking approach as link farming and the business's revenue fell to zero. Working with software engineers and seeing how they used Jira gave him the idea for a project-management tool for M&A. DealRoom later focused on buyers, including large companies and private equity-backed roll-ups, and expanded across pipeline management, diligence, integration, and reporting.

### Customer interviews changed DealRoom from a marketplace idea into a focused product
[19:00](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1140s)
DealRoom initially tried a marketplace model and launched with about 1,300 users and 200 deals. Patel concluded that it had created a large collection of poor-quality opportunities rather than a useful market. The company went back to customer development interviews and learned to separate founders' attachment to an idea from the problem customers actually experience. Patel recommends open questions, listening for repeated patterns, and testing different customer groups. He prefers around 40 conversations because five or six calls may not reveal patterns across prospects.

### Contract analysis became the first AI use case because customers could connect it to legal savings
[18:56](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1136s)
When DealRoom explored AI, Patel had several ideas about where it might help M&A. Customer discussions showed that the clearest value came from reducing legal expenses, so the team started with contract extraction and analysis. A buyer may need to review hundreds of employment, customer, and vendor contracts for change-of-control rights, assignment provisions, or consent requirements. AI can turn this unstructured material into structured information and explain what it found. Patel says this use case won support because its value was easy to connect to a real invoice, unlike some of his more ambitious ideas.

### A small product team helped AI development move faster inside a mature company
[20:20](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1220s)
Patel says the product-development habits that work in a two-to-five-person startup do not work unchanged in a company of about 50 people. Engineering time is divided among existing customer requests, sales requests, and new product work, with only a small share reaching genuinely innovative features. DealRoom moved the AI roadmap into FirmRoom, a separate data-room product that had already reached about $1.5 million in revenue with little marketing or product investment. The AI team could move faster because it had fewer competing priorities, while still benefiting from DealRoom's existing platform, security credentials, customer base, and market reputation.

### AI pricing becomes harder when a new feature is added to an established product
[31:59](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1919s)
Patel does not have one pricing model for AI. He says companies should consider established pricing norms when they exist, then compete through product differences instead of automatically undercutting the market. Usage-based pricing can fit some infrastructure products, but it is harder for an end-user product with a simple subscription expectation. DealRoom's sales-led product can sell AI through document or credit tiers, while its self-service product needs pricing that is transparent and simple. Patel describes a flat monthly fee as one possible fit inside the existing model, while acknowledging that heavy document use could create cost problems.

### Distribution becomes more important after a company has learned to build product
[30:03](https://www.youtube.com/watch?v=1QzijeBH7TU&t=1803s)
Patel agrees that some AI companies reach revenue with small teams because they are product-led, accept credit cards, and benefit from virality. He also says many such companies struggle with distribution, brand building, and selling to conservative industries. DealRoom benefits from its existing position in M&A, and Patel says its podcast and social media now drive about half of its new business. If he were starting in a new industry, he would think about distribution earlier rather than focusing almost entirely on the product.

### AI hiring requires defining the actual gap instead of copying broad job titles
[42:00](https://www.youtube.com/watch?v=1QzijeBH7TU&t=2520s)
Patel left the detailed AI team design to DealRoom's VP of engineering because AI job categories are becoming difficult to separate. His advice is to define the business problem, identify the specific skills needed, and write a job description around that need. He makes a similar point about design hiring, where DealRoom needed stronger UX work while still requiring enough UI ability. Patel also connects company growth to communication, recognition, and a workplace where people feel comfortable speaking with colleagues. He says organizational psychology books taught him more about leadership than leadership books did.

### Buyer-led M&A works better when acquisitions follow a defined strategy
[48:04](https://www.youtube.com/watch?v=1QzijeBH7TU&t=2884s)
Patel supports the idea that companies are bought rather than sold when the buyer has a clear strategy and actively identifies suitable targets. He gives geographic expansion as an example: DealRoom may want an established European company with a similar customer base because buying that business could be faster than entering Europe from scratch. A defined target profile lets the buyer explain why the companies fit and how they could create value together. Patel contrasts this with an adviser-led sale, where compressed timelines can reduce diligence and leave buyers with integration assumptions that do not hold.

### AI acquisition interest is constrained by weak defensibility and high expectations
[52:43](https://www.youtube.com/watch?v=1QzijeBH7TU&t=3163s)
Patel sees a hype cycle around AI M&A. Some startups have little defensible intellectual property beyond prompt instructions, while companies with proprietary data can receive stronger valuations. Venture investors may also have high exit expectations because of the prices paid in the funding market. Patel does not expect AI companies to be acquired indiscriminately. He predicts a smaller group of winners, many failures, and a correction in how these businesses are valued. He also expects different AI technologies to receive different valuations based on the strength of their intellectual property.

## Notable quotes
- Kison Patel: "The one thing from that experience, it was the first time I was working with software engineers. I was really intrigued how they were using Jira to manage developing our software platform and that led to this inspiration of why not build a project management tool for M&A." (18:07)
- Kison Patel: "I think this is the most valuable step in the entrepreneurial journey is validating that you're solving the right problem." (19:38)
- Kison Patel: "The number one thing that we found was people wanted to save legal expenses." (20:19)
- Kison Patel: "I think you're going to see few winners and a lot of losers in the coming years and probably a real correction around how these companies are valued." (53:56)

## Tools & references mentioned
- DealRoom
- M&A Science
- Jira
- Google
- Running Lean
- The Mom Test
- FirmRoom
- Slack AI
- AOL
- MLOps Community
- Deals from Hell: M&A Lessons that Rise Above the Ashes
- Whiz

## Who should watch
- You are adding AI features to an established B2B product and need a way to choose use cases that customers will pay for.
- You are deciding between usage-based and simple subscription pricing for an AI capability with uncertain model costs.
- You work on acquisitions and want to connect target selection, diligence, integration, and company strategy instead of reacting to businesses put up for sale.

## Related talks

- [Anatomy of a Software 3.0 Company](https://mlopstalks.com/talks/anatomy-of-a-software-3-0-company) (Sarah Guo, Conviction, 35:21)
- [AI in Education Fireside Chat](https://mlopstalks.com/talks/ai-in-education-fireside-chat) (Klinton Bicknell, Duolingo & Bill Salak, Brainly & Yeva Hyusyan, SoloLearn, 31:01)
- [AI's Next Frontier](https://mlopstalks.com/talks/ais-next-frontier) (Aditya Naganath, Kleiner Perkins, 56:04)
- [From A Coding Startup to AI Development in the Enterprise](https://mlopstalks.com/talks/from-a-coding-startup-to-ai-development-in-the-enterprise) (Ryan Carson, Intel, 58:16)
- [Building AI Products across Multiple Domains: Commonalities & Non-Commonalities](https://mlopstalks.com/talks/building-ai-products-across-multiple-domains-commonalities-non-commonalities) (Dhruv Ghulati, 12:32)