The AI line in your underwriting: How are you calculating that ROI?

 profile

June 02, 2026

by a searcher from University of Texas in Folsom, CA, USA

If you're underwriting a deal right now, there's a decent chance your model has a line for AI driving down costs or expanding margins. Some recent observations have caused me to re-examine this line. The ROI is real but there are some important nuances. Curious to hear others thoughts on this too. Here's the pattern I keep thinking about: AI runs at the speed of your worst process, your dirtiest data, and your least willing employee. The models are incredibly capable and relatively cheap now. The ROI is gated by everything around it. I'm seeing this from inside my own company. I run a profitable software business for searchers, and a good chunk of the daily work is done by AI agents while I review instead of type. I had it easy: it's greenfield, no legacy process to redesign, no 20-year-old data to clean. And even then, the wins came from wrapping each agent in a real workflow and pulling the parts that have to be exact into actual code. A couple weeks ago I walked a group of independent sponsors through this. The interest was real, a few emailed me before I'd packed up my laptop. The people buying real companies are actively trying to figure out where AI fits. So here's where I have been looking first based on my daily experience with agents. The high-volume, rules-based, time-sucking tasks. The repetitive stuff a sharp associate quietly hates doing. An agent wrapped in a clear process can pay for itself fast there. The judgment calls and the messy, undocumented work come later, after you've done the unglamorous part: writing the process down and cleaning the data so the agent has something solid to stand on. I'll admit I've gotten pulled in. A few companies have asked me to build agents for them, and between that and running the software business, my own search has taken a back seat. I'm okay with that. Applying AI to real businesses is the most interesting work I'm doing right now. None of this is settled yet. I'm a sample size of one. For those of you who've deployed AI in a business you own or operate: where did the ROI actually show up, versus where you expected it? Also, thanks @Christian Berger‌ for hosting me at McGuire Woods independent sponsor luncheon.
0
2
43
Replies
2
commentor profile
Reply by a searcher
from University of Southern California in Fremont, CA, USA
In my experience, building automated Agents/scripts for existing processes is the toughest. In a lot of businesses/ companies a lot of processes or AI automatable work is almost always a manual process. And human behavior to protect the turf/role always trumps. On the other hand like you mentioned, it is easier to do this in an almost green field environment where one does not have to fight the existing systems and people's roles. I have seen immediate ROI with the list in the backlog which unlocks revenue versus existing processes. The ROI in cleanup/automation for operational efficiency with existing processes is always a long term goal.
commentor profile
Reply by a searcher
from University of Texas in Folsom, CA, USA
Agree with your thoughts here Anjana. My post didn't even talk about the people and change management component. So true. I've led large transformations as an ops exec or as a Deloitte consultant and getting people to accept change is one of the hardest parts of driving ROI in technology improvements.
Join the discussion