AI Prospect Discovery vs an In-House Team: The Real Cost Comparison

Software versus analysts is the wrong comparison. What is actually published, what is not, and how to build the number for your own firm.

The pitch for AI prospect discovery is a cost argument: software costs less than analysts, therefore replace analysts with software. It is a bad argument, and not because software is expensive. It is a bad argument because it compares two things that do not do the same job.

This article sets out what each side of that comparison actually costs, what can be verified and what cannot, and where the saving genuinely is.

Start with what is actually published

Almost nothing in this market has a public price. Grata publishes no pricing and quotes on request. SourceScrub publishes named tiers, Essentials, Plus and Professional, with no figures attached. Cognism quotes only. DealCloud has no list price, and the estimates circulating on procurement comparison sites are third-party guesses that differ from each other by an order of magnitude.

The clean exception is Apollo.io, a sales intelligence tool rather than an M&A one, which publishes self-serve pricing from zero to 149 dollars per user per month.

We will not publish a headline "software costs X, an analyst costs Y" table, because building one would require inventing at least one of the two numbers. Anyone who does publish that table has either got a quote you cannot see, or made it up.

If you want defensible salary benchmarks for the other side of the comparison, the citable sources are the Heidrick and Struggles 2025 North America Private Equity Investment Professional Compensation Survey and the eFinancialCareers 2025 pay research. Use those directly rather than a vendor's summary of them.

The comparison is wrong anyway

Set the numbers aside. The structural problem with cost-per-prospect thinking is that it prices the cheap part of origination.

Building a list is the cheap part. It was cheap when an analyst did it with a trade directory and it is cheaper now. What costs money is everything after the list.

Deciding which twenty of two hundred companies are genuinely worth pursuing, which requires understanding why service revenue is worth more than installation revenue in building services, or why a care home's occupancy trend matters more than last year's margin.

Making contact in a way that does not get ignored. Founder-owned businesses in fragmented trades receive a lot of unsolicited approaches and most are deleted.

Holding a relationship open for eighteen months until the owner is actually ready, because the moment of readiness is rarely the moment you called.

Being the party the owner calls first when readiness arrives. That is a relationship outcome, not a data outcome.

An AI tool compresses the first stage substantially and does very little to the rest. If your origination cost is dominated by the later stages, and in most acquisitive businesses it is, then the software saving is real but small in proportion to the total.

What the industry research actually says

There is credible published work here, and it is worth reading in the original rather than through a vendor's marketing.

Datasite, with FT Longitude, surveyed 1,000 dealmakers and published the findings on 24 June 2026. The headline was that most dealmakers no longer consider human-only decision making defensible in complex deals. Read carefully, that is a statement about augmenting judgement, not about removing headcount, and the study does not claim AI identifies willing sellers.

Bain's Global Private Equity Report 2025 includes a section specifically on generative AI adoption in the industry. Preqin's 2025 Global Report on private equity and its Deal Flow Monitor Outlook 2025 cover deal flow conditions. SS&C Intralinks has published both an AI in M&A dealmaking benchmark study and its H2 2026 Global M&A Dealmakers Sentiment Survey.

Between them these give you a defensible view of adoption. None of them, as far as we can find, publishes a validated cost-per-sourced-deal comparison between software and human teams. If your investment committee wants that number, it will have to be built from your own pipeline data.

How to build the number for your own firm

This is the calculation that is actually worth doing, and it takes an afternoon.

Take the last three completed acquisitions. For each, write down where the opportunity originated: an inbound approach, an existing relationship, a broker, a cold outreach campaign, or a database search. Then estimate the hours spent between first contact and heads of terms.

Most acquirers who do this discover something uncomfortable. The deals that completed came from relationships and intermediaries. The database-sourced names generated activity, meetings and a great deal of pipeline reporting, and rarely converted.

That does not mean the software was wasted. It means it should be priced as a research and coverage tool, which is worth having, rather than as an origination engine, which it is not.

The honest conclusion

Buy the tool if it saves your team research hours, and it probably will. Do not buy it on the promise that it replaces a person, because the person's value is concentrated in the part the tool does not touch.

And keep a second channel open. The businesses worth buying in fragmented trades are usually not on any market, and their owners talk to a sector adviser before they talk to a buyer.

That is what we do. We act sell-side in building services and facilities management, in testing, inspection and certification, and in specialist healthcare, in the UK and the United States, and our mandates are off-market. Register your acquisition criteria below and we will approach you when a mandate matches rather than adding you to a mailing list. There is also the off-market register and the current acquisition briefs.