Full episode transcript: AI & PE: The Future of Value Creation

Guest: Ash Pembroke, Managing Director, AI Strategy and Value Creation, Accordion | Host: Kyle Roemer

Ash Pembroke: If you can build a tool set that can make your fastest horses faster, I think that’s a much more compelling and clearer value proposition than trying to bring the bottom 30% of the workforce to parity with the middle 50%.

Kyle Roemer: Welcome to AI & PE: The Future of Value Creation — the show where we cut through the noise to hear directly from experts, practitioners, and members of the PE community on how AI is transforming portfolio companies, accelerating growth, and rewriting the playbook of private equity. I’m your host, Kyle Roemer, Managing Director and Head of US Data and Analytics here at Accordion. In today’s episode, we’re joined by Ash Pembroke, Managing Director of AI Strategy and Value Creation at Accordion. She’s got great advice on how to approach your AI strategy — from accelerating your strongest performers, to pushing your leadership team to identify your biggest business needs, to where AI, data, and tech deliver the highest ROI. Here’s my conversation with Accordion’s Ash Pembroke.

I thought I’d just start with — what’s the biggest “wow” moment you’ve had so far in AI, in your career?

Ash Pembroke: I was with a co-worker at a former firm, and we’d challenged ourselves to grow without increasing headcount in functions where you’d normally assume you needed to — sales, sales ops, engineering. I saw a mid-sized firm, with really good process hygiene and basic gen AI integrations between Slack, Salesforce, and some custom development, scale their team — going from $25 to $50 to $100 million — without materially increasing headcount. They invested in caliber, not necessarily in supporting personnel.

That was the first “aha” moment for me — seeing all these tools come together in a function to drive a big-ticket item for the leadership team, and getting to be part of it. It converted me to the idea that AI could help mid-market companies scale and grow really quickly, and I became genuinely passionate about it.

Kyle Roemer: We’re constantly talking today about how service businesses are getting disrupted. How are upstarts catching up to more established players?

Ash Pembroke: It’s getting harder and harder to tell how many people actually work at a company, or how quickly they’re turning around proposals. Smaller shops are able to show up at scale in a way I don’t think they could have before. And even if their hit rate on lead conversion stays the same, the number of at-bats they get is a lot higher. That’s a really interesting, and challenging, thing to consider.

Kyle Roemer: What’s an example — in service delivery or go-to-market — where you really saw the impact, where you didn’t have to think about SG&A scaling the same way you might have historically?

Ash Pembroke: I’ll give you a concrete one. I’ve been working with a customer that’s a roll-up of HVAC, plumbing, and electrical businesses. Labor turnover is a huge issue there — and it’s not just turnover of bodies, it’s turnover of less experienced people who misestimate jobs and mess up work. There’s a cascading effect when experience leaves.

One of the most compelling uses of AI in the field: an app trained on essentially every version of HVAC equipment from any OEM. All an inexperienced technician has to do is snap a picture, and it automatically catalogs the system they’re looking at and surfaces knowledgeable estimates. It makes someone new to the job much more efficient, faster — and the accuracy of every downstream decision starts to improve.

At the enterprise level, I have mentors at large OEMs eating technical debt for breakfast — in their automotive platforms, legacy codebases, hardware structure, it’s public, it’s in the news. A few years ago, if you’d asked what it would cost to drive an enterprise transformation program at one of these large companies, you’d have seen price tags in the hundreds of millions, with Sisyphean timelines. That’s getting much crisper now with the help of gen AI and code development. It’s becoming material — error rates are going down. As a former Luddite, over the last 16 months or so, I’ve become much more of a believer.

Kyle Roemer: I fully agree. The idea of saving a single headcount in a department — we shouldn’t even be talking about that anymore, frankly. Every single person, in every department, should be leveraging these tools to be more efficient and effective.

Ash Pembroke: Take a consulting business like ours — you have this interface with the customer: someone enters a lead into a CRM, then logs meeting notes, then does this, that — you’re context-switching across multiple applications the whole time. In the past, if you did a time study, you might throw RPA at that, or some random daisy chain of solutions. I’ve seen almost all of that automated away with custom integrations across Slack, Salesforce, whatever your contract repository is — I can spin up a lead in DealCloud or Salesforce, memorialize a conversation into notes, turn those notes into bullet points for a proposal, all within the console of something like Slack. There are plenty of AI-native businesses doing exactly that.

Kyle Roemer: There’s an interesting dilemma I keep running into — are you automating away the substance and the expertise? If you’re pitching a client in professional services, are you losing something there, or are the models good enough now that it’s providing the right level?

Ash Pembroke: I have a more forgiving way of evaluating how AI helps me. How long would it take me to explain what the customer said, and what I think the shape of the proposal should be, to a third-year associate? How long would that take, and how many iteration cycles would we go through? That’s what people are grappling with across the economy, if you zoom out — the structure of how we as knowledge workers build and craft our teams.

A well-tuned enterprise GPT, with agents that can plug into creating a PowerPoint or a proposal, can get you as far as a first draft. Is it perfect? No — but it wouldn’t have been perfect either way, because so much of our work is caught up in data-entry drudgery, or translating a meeting into notes, or notes into a CRM entry. Every team experiences this “swivel chair” problem, and if we’re honest, it’s one of our least favorite parts of the job. What if it just didn’t have to be?

Kyle Roemer: I like that framing — if it’s painful, manual, and boring, shouldn’t we be thinking about AI for it? Shifting gears a bit — we spend a lot of time with PE-backed CFOs and sponsors. If you’re one of those sponsors, putting half a million or a million dollars into next year’s CapEx budget — whether you’re in services or industrials — what are you putting it toward? Where’s the return?

Ash Pembroke: I’m hearing a lot of “we have to be AI-enabled at exit — what does that mean for us, what have we invested in so we can tell that value story?” A couple of areas AI opens up: there’s traditionally been reticence to operationally integrate certain functions — say, getting three or four product companies to sell together, different sales forces, rebranding, creating a combined services catalog with a pricing strategy baked in. That normally takes a year or two to get people “walking and talking” the new way before you start recognizing value. AI-enabled learning management solutions can scale people’s change-readiness and adoption of new operational processes much faster — I think that’s a huge, largely untapped opportunity.

AI is built on data. A lot of us have looked at the value creation plan and said, “I’m not going to invest a million dollars in a data warehouse right now — group A has this data mart, sales ops has that, we’re all fine in our own silo.” Then you miss the exit timeline, or the plan changes, and suddenly you need to incentivize different purchasing behavior or shift lifetime customer value — and you can’t do that without the data. What’s compelling now is that harmonizing that data, talking to all the stakeholders, designing the data catalog and the EDW with AI-enabled teams doing the data engineering and requirements gathering — that’s months, not years, at enterprise scale. We should be challenging ourselves to do data work faster and more elegantly than we have in the past.

Kyle Roemer: You need to do that work — yesterday, ideally. Just be clear you won’t get a ton of direct value from it alone, since it’s your entry point into AI. It’s the second and third use cases that take a bit more risk and redesign something core to the business.

Ash Pembroke: The last category I’d point to: I was with a group of CFOs this morning, and I told them — go back and look at the last three years of strategy, at every place you said “it costs too much,” “we’re not qualified to move into that adjacent market,” “I’d need to spin up another team.” Go back and assume every step of that business case is now enabled — the denominator has changed. We have to reevaluate those strategies. I see teams pulling ahead by thinking creatively there and executing on it, because that’s real growth, not just cost takeout — removing one headcount from a shared service function is small potatoes relative to what sponsors actually want. Even if it feels a little uncomfortable, challenging ourselves to think big right now, instead of just tactical value creation, is a good exercise on its own.

Kyle Roemer: Go back to the value creation plan, push it through the paces of where the technology is today, and think about what you’d do differently on product, service, or market expansion. Massive, underutilized opportunity right now. What’s a fundamental thing CFOs could be doing with AI today that you think is being overlooked?

Ash Pembroke: I had a long conversation with a leader who said he spends most of the month closing the books, and then another week compiling datasets for his board deck — there’s not a single week he isn’t closing the books, prepping for a board meeting, or working the weekend to add commentary on why a number moved. If it involves routine commentary generation on data that just goes up or down every month, start using AI not just to create the content but to auto-summarize it for you. I’ve yet to see a first draft of a CFO deck that doesn’t get entirely rewritten by that person on a Saturday morning anyway — so let the AI output be a little imperfect and let it fast-track things: accelerating the close, auto-generating commentary, using off-the-shelf computer vision to extract text from PDFs for competitive intelligence. Look at everywhere your team is struggling with time management because of process-driven work, get those pieces optimized so you can actually breathe — and once you can breathe, you can go answer the secondary and tertiary questions, or test something new in market, instead of just keeping the lights on.

Kyle Roemer: Let’s talk about something that’s been all over the headlines the last few weeks — the MIT study finding that 95% of AI projects or pilots are failing. What’s your advice to a CFO or CTO on incrementally improving their success rate, versus other technology rollouts they’ve tried in the past?

Ash Pembroke: Part of why things fail is we take a tool that does one thing really well, get it to parity across the workforce, and roll it out broadly — meanwhile your top 25% of performers are saying, “this isn’t useful to me, I’m already three times more efficient than the bottom 25%.” When you pilot this stuff, you’ll get an adoption group that’s excited, and another group that’s more resistant and says “this is stupid, I don’t need it.”

I’d encourage you to build a tool set that makes your fastest horses faster — that’s a much more compelling value proposition than trying to bring the bottom 30% of the workforce to parity with the middle 50%. Mobilizing your highest performers at a tactical level, and empowering them to design solutions, is where I see the biggest differences.

Kyle Roemer: Can you make that real for the audience — what does that actually look like in AI work?

Ash Pembroke: Here’s a tactical example. At a former firm, we rolled out ticket-routing functionality — a contact-deflection business case for a mid-sized SaaS company that had moved to an offshore shared service center, hadn’t documented well, and had heavy turnover in the newest 30% of their CSRs. They saw a material productivity increase from that alone.

What they didn’t do: take their most productive people and give them more autonomy — more decision-making power and data within their role, similar to the Four Seasons model, where an employee gets a certain amount of discretionary latitude per day. If you couple that with productivity tools that get your highest performers out of drudgery, you might not need the bottom 30% of the pyramid at all. When these pilots show great metrics but wash out at scale, it’s usually because a lot of the workforce didn’t actually need the tool — leadership was grasping at whatever looked like the biggest problem, and the easiest thing to automate first usually isn’t the biggest problem.

Kyle Roemer: Definitely an exciting time to move past that first-level use case into something more transformative.

Ash Pembroke: The flip side is when the problem genuinely can’t be solved by humans at that scale. In that same SaaS company, what they actually needed wasn’t automating the first few steps of a CSR interaction — it was a customer-success onboarding agent that could walk a newly onboarded company through every configuration step, SLA, and setting. That’s what got them the biggest lift, and it also standardized their data at onboarding. It created efficiency across the whole value chain — but it took longer, because leadership initially wanted the quick, safe win instead.

Kyle Roemer: For a while now — the last 12 to 18 months, certainly — there’s been a lot of small efficiency plays with people taking very small risks, and correspondingly small returns.

Ash Pembroke: The safe use cases are getting commoditized into tooling now — turn it on, use it, but then move to phase two. That’s where you’re seeing a lot of quietly transformative pilots, because people don’t want to talk publicly about what they’re building since it’s genuine IP.

Kyle Roemer: These have been great, real examples from the market. What ingredients should PE-backed businesses look for in a partner to deliver real value with AI?

Ash Pembroke: I’d challenge everyone to think about solving problems with small teams — you don’t need an army of people to write documentation, transcribe meeting notes, or turn a meeting into a spreadsheet. Think about all those artifacts that just help you get to a decision. I’m more confident than ever that a small group of people with interdisciplinary skills and the right tools can execute real strategic change without over-engineering the size and scope of it. As an introvert myself — we don’t need twenty people at this party, three or four is plenty. Good for the budget too.

Kyle Roemer: Not unlike the “two-pizza team” concept — though now you could argue it should be a one-pizza team. A three-to-five person team delivering massive impact, as long as they have multidisciplinary skills to tackle the problem.

Ash Pembroke: I’d say hire people who think in patterns and repeatable methods — scientific method, essentially: how would I know this is true, not just conjecture, but also comfort with “I’m pretty sure, here’s what I don’t know yet, and here’s how I’ll figure it out.” People with an owner mentality. What’s genuinely challenging, even in how we recruit, is that I’m essentially asking people to look at their own job and automate away the drudgery in it — which can feel a little threatening at first, and it also pushes people to move laterally across functions, since they can’t just do one thing well anymore. That pressure-tests traditional job descriptions. Somebody comfortable with a certain amount of ambiguity is really important here — if you’ve got people like that in your organization, I’d be pulling them out and empowering them.

Kyle Roemer: That’s it for this episode of AI & PE: The Future of Value Creation. Be sure to follow the show on your podcast app of choice so you don’t miss our next conversation on how AI is reshaping private equity. Follow Accordion on LinkedIn for more insights on value creation and private equity, and if you want to learn more about what we do, visit accordion.com.