EisnerAmper

Matthew Morgan is a Partner and the COO of EisnerAmper, a professional services firm with 4,500 professionals. He previously held roles at McKinsey & Company and Bridgewater Associates and served as an officer in the U.S. Army.

In the span of a single week this February, I found myself in two conversations that I suspect will shape how I think about leadership and technology for years to come. The first was at Intapp Amplify 2026 in New York, where a group of practitioners from professional services and financial services gathered to take stock of where AI is actually landing in their organizations. The second was at Microsoft's Times Square office, where a smaller forum brought together business leaders to grapple with what the current AI moment demands of the people running enterprises.

I went to both events as a learner. I left with more questions than I arrived with which I've come to believe is the surest sign that a conversation was worth having.

The context for Amplify was hard to ignore. In early February 2026, Anthropic's launch of Claude Cowork erased roughly $285 billion in software market capitalization in a single trading day. Within weeks, Adobe, Microsoft, Salesforce, SAP, ServiceNow and Oracle had collectively shed more than $730 billion in value. Traders at Jefferies coined the term SaaSpocalypse. The IGV ETF, which tracks North American software companies, fell close to 30 percent off its September peak — back to valuation levels not seen since 2014. The narrative in the financial press was stark: GenAI was coming for SaaS, and the incumbents were running out of time.

What I Heard in the Room Told a More Complicated Story.

Darren Cohen of Goldman Sachs Global Growth Equity suggested the SaaS versus GenAI framing may be a false dichotomy. Rather than competing forces, the two are more naturally understood as complementary ones. SaaS brings what regulated industries rely on most — identity management, systems of record, defined business processes, compliance-oriented workflows. These aren't constraints waiting to be disrupted. They may be precisely the foundation on which GenAI can operate most responsibly and effectively.

Alex Chiou of HPS offered a counterweight to the surrounding hype: the organizations most likely to benefit from AI are those starting with a clear problem statement rather than a technology in search of an application. In a landscape saturated with both enthusiasm and panic, that kind of discipline seems increasingly rare and increasingly valuable.

One framing stayed with me: structured versus unstructured processes. SaaS tends to excel at governing structured workflows. GenAI may find its greatest contribution in the unstructured spaces synthesizing information, surfacing patterns, accelerating judgment calls, where rigid systems have traditionally struggled. If that's right, the two aren't substitutes so much as complements covering different parts of the same terrain.

The unveiling of Celeste, Intapp's new AI interface developed in partnership with Anthropic, felt like an early illustration of this in practice: not a replacement for a system of record, but an intelligent layer built around it. Whether the orchestration model ultimately prevails over displacement is still an open question. But the markets may be pricing in a scenario that practitioners on the ground aren't actually living.

 ​When smart people in good rooms are telling you something is coming, the question is no longer whether to believe them. The question is whether you're imagining the pace correctly. 

This was not, I should say, the first time I'd found myself in a room where the AI conversation felt ahead of where most organizations actually were.

In 2021, I participated in a panel at Intapp DealCloud's Amplify event with Nigel Riley, Amanda Tooker and Jim Brady, discussing how professional services firms could harness client data to move beyond traditional risk management. My argument then was that unlocking AI's potential required moving off email-based, manual processes onto enterprise platforms that could connect disparate data sources. The "adverse media monitoring" I described as a promising capability was still largely periodic and human-curated.

I Got the Direction Right. I Underestimated the Speed by a Significant Margin.

The move from unstructured to structured data that I described in 2021 as a strategic aspiration is now the price of admission to an entirely new generation of AI capabilities. Firms that did that foundational work are able to automate, predict and act in ways that were science fiction four years ago. Firms that didn't are finding themselves locked out, not because they made the wrong strategic call, but because the timeline compressed in ways none of us fully anticipated.

What I carry from that experience is a particular kind of epistemic humility. When smart people in good rooms are telling you something is coming, the question is no longer whether to believe them. The question is whether you're imagining the pace correctly.

Which Brings me to the Microsoft Forum, And to the Two Perspectives that Landed Hardest.

Allie K. Miller opened with a provocation: AI is not a tool. Tools are picked up and put down. What she was describing is something categorically different — a transformation of how work itself is organized. Her example was vivid: Walmart invited 100 vendors to negotiate contracts with an AI agent. Eighty-nine accepted. Of those, 68 percent closed deals. The agent did not assist the negotiation. It conducted it.

Matthew Loys Duncan of Microsoft's WorkLab research introduced a concept I haven't stopped thinking about the Frontier Firm. His definition — a human-led, agent-operated organization that buys intelligence like electricity, puts it to work like an employee, and compounds it like interest — is the kind of framing that clarifies something you already sensed but hadn't quite articulated. His central argument: capability is no longer the bottleneck. The constraint is leadership imagination and prioritization.

Most organizations are still running on org charts with fixed hierarchies designed for a world where humans directed and performed all knowledge work. The Frontier Firm replaces the org chart with a work chart: a fluid structure built around outcomes, drawing on the right mix of humans and agents to accomplish them.

I Left Both Conversations Carrying a Question I Don't Yet Have a Satisfying Answer To.

The SaaS debate, I think, will resolve itself, probably in the direction the Amplify conversation suggested, with AI and enterprise software finding a complementary equilibrium rather than a winner-take-all outcome. That's a technology and market question and smart people are working on it.

The harder question is organizational. If capability is genuinely no longer the bottleneck, if the constraint is leadership imagination, then what does it mean to lead well in this environment? What decisions are being made by default, in the absence of a deliberate answer? And how many organizations are operating with an org chart designed for a world that no longer fully exists?

At EisnerAmper, these are the questions I'm actively working through. I don't think the answers are obvious, and I'm skeptical of anyone who suggests they are. What I do believe is that the leaders who will navigate this well are the ones willing to stay genuinely curious, to keep walking into rooms, asking hard questions and resisting the pull toward premature certainty.

Two rooms in February reminded me of that. I expect the next few years will offer many more.