Harry Ratcliff

Founder | AI implementation for the mid-market
Share learnings at IB | 4 kids, 1 wife

[Harry Ratcliff @hgr_la]

Hi, I'm Harry. I'm the co-founder and CEO of DealSage (https://t.co/ibdU9ammka).

Before this, I spent 10 years in investment banking at J.P. Morgan advising boards on M&A and capital allocation. My co-founder Ben Amor spent a decade deploying AI at Palantir.

Our goal is to take the Palantir playbook and bring it to the middle market, specializing in private equity-backed companies and the firms behind them.

To become AI native and make better decisions, you need to understand your business. And to understand your business, you need to aggregate and understand your data.

That is what we build at DealSage:

- Connect: We pull your fragmented ERP, CRM, billing, and operational systems together.

- Centralize: We organize that raw data into one unified company model so every department works off the same truth.

- Implement: Our forward-deployed engineers sit inside your operation to build custom AI workflows and agents around how your team actually works.

- Decide: Your leadership team gets real-time visibility and automated execution that drives measurable EBITDA.

On here, I write about what actually works (and what breaks) when you deploy AI inside real operating companies, plus a weekly Sunday newsletter on how AI is reshaping business models.

For more info or to book a call, check out https://t.co/ibdU9ammka.

[Harry Ratcliff @hgr_la]

Finance was using contract terms the customer had already changed.

We found this reviewing a services contract with a PE-backed operator as part of our AI work. The customer had waived a performance requirement by email, but nobody had drawn up an amendment.

The PDF still looked like the official answer. The change was sitting in an inbox, and the finance model was working from the original terms.

We treated the email waiver as an amendment, filed it against the contract and showed the live terms alongside the signed ones on a scorecard.

That gave finance a way to see what the business was actually working to, with the correspondence there to check.

[Harry Ratcliff @hgr_la]

@justinskycak Or doesnt even need to be automated in the first place

[Harry Ratcliff @hgr_la]

@unusual_whales Then you layer on top all the downstream impact as well (Data centre infra, electricians and servicing, electrical grid) and you realise that everything is AI now and any weakness will result in a complete economic collapse.

[Harry Ratcliff @hgr_la]

I've gone full circle on all the "Sales best practices" and now basically think all these tactics and tips are complete bs. If you write a good email thats relevant then people will reply if they're interested.

Any time i see "I'm curious if..." or "Worth a conversation..." it just instantly goes into the bin now.

[Harry Ratcliff @hgr_la]

Not sure about this, maybe executing the trade sure but the whole point of trading desks is that you're the market maker, so what you think and are willing to transact at dictates the market itself.

Eventually sure but anyone thats seen agents trade stock portfolios will know they just regress to the mean and underperform.

Basically markets don't trade rationally whereas agents do.

[Harry Ratcliff @hgr_la]

@dave_stickland Its true, i think your brain kind of tricks you into thinking it was that way a bit too, like cover up all of the pain

[Harry Ratcliff @hgr_la]

@AizikZimerman Has this worked though? Do you see better alignemnt / more estimates made etc?

[Harry Ratcliff @hgr_la]

@ndbroadbent Do you find cold email is still effective?

[Harry Ratcliff @hgr_la]

@signulll Same reason theres a big fight to be your notetaker, see Granola, Wispr etc

[Harry Ratcliff @hgr_la]

I like being able to check an answer another way.

We implement AI in PE-backed businesses, and on a field services project we used truck GPS to work out time on site, then checked it against a second, simpler method.

Both gave the same answer. A bit easier for the operating team to act on that than just take our word for it.

[Harry Ratcliff @hgr_la]

@IlliquidInsight I'm curious/concerned to see whether/how much this exacerbates over the next 5 years. I can't look past how rising cost of capital from AI drain isn't going to massively impact the LMM.

[Harry Ratcliff @hgr_la]

Biggest AI misstep right now is focusing on cost cutting / automation and "time savings".

You have to be more ambitious now, what can we take from our data that allows us to understand our business better, and then use that to address the real bottlenecks thats stopping us from achieving our goals.

[Harry Ratcliff @hgr_la]

@unusual_whales I'll believe it when i see it

[Harry Ratcliff @hgr_la]

@PEoperator Looking across X i think thats already happened...

[Harry Ratcliff @hgr_la]

@AizikZimerman In all facets of life and business the goal should be to reduce, reduce, reduce. Simplify and try to remove as much unnecessary work as possible.

[Harry Ratcliff @hgr_la]

@signulll I want a new "phone" in general, i dont want to be tapping around, i just want to talk to it and things happen. Its surprising to me that this isn't obvious to Apple

[Harry Ratcliff @hgr_la]

I'd start with the annoying bit of the morning.

When we're discussing AI implementation with a PE-backed company, I want to understand what gets in the way of someone doing their job.

- A manual check they have to repeat before getting started
- Information they keep moving between systems
- Work that holds up colleagues further down the process

That gives us something concrete to build around, inside tools the team already uses. It's an easier conversation than asking everyone for their AI strategy.

[Harry Ratcliff @hgr_la]

AI adoption reports can leave out the useful work.

When we discuss implementation with PE-backed companies, I want to understand what the team is getting through. The measurements can drift towards whatever is easiest to count.

The mistakes I'd watch for:

1. Treating bought seats as progress. Access tells you very little about what has changed in the operating day.

2. Making weekly active users the main scorecard. A person can benefit from AI inside their normal tools without opening a chat window.

3. Measuring a task in isolation. The manual checks around it and the next handoff still affect whether the work gets through.

4. Losing sight of the operating result. The workflow should have been chosen because the business wanted something to improve.

I'd name that workflow and the expected result at the start. It makes the conversation about whether the rollout is helping much easier later.

[Harry Ratcliff @hgr_la]

@spikeballchris 😂 oh and dont forget your 8 hours of sleep a night too, that ones super important.