AI won’t just transform factories and logistics. It’s coming for lawyers, teachers, and accountants

“AI won’t transform the jobs we always assumed were vulnerable, it will transform the ones we thought were safe,” said Shay Grinfeld, Managing Partner, and Josh Trup, Investor at Greenfield Partners. “Lawyers, teachers, accountants, consultants: software never touched them because they were too human, too judgment-driven. But voice, vision, and reasoning combined don’t automate tasks, they replicate judgment. That’s the real disruption - AI is coming for the professions that defined the last century. A big blue ocean to build.”

They joined CTech to discuss how factories, warehouses, and logistics may have seemed like the natural first targets for AI disruption, but that the bigger shock is coming elsewhere.

Shay Grinfeld, Managing Partner, and Josh Trup, Investor at Greenfield Partners

You can learn more in the interview below.

Fund ID
Name and Title: Shay Grinfeld
Fund Name: Greenfield Partners
Founding Team: Shay Grinfeld & Yuda Doron
Founding Year: 2016
Investment Stage: Early-Growth & Growth
Investment Sectors: AI, Cyber, Infrastructure tech, Deep-Tech, Vertical AI, Defense

On a scale of 1 to 10, how has AI impacted your fund’s operations over the past year - specifically in terms of the day-to-day work of the fund's partners and team members?

9 - AI is now fused into how we run the fund, and its greatest dividend has been giving us back time. Time we can now spend with founders, time to think more deeply, and time to make better, more context-aware decisions at speed. The paradox is that the cost of information has collapsed, so what used to be an edge is now a commodity (e.g. data signals, market research, etc.), pushing us all to focus on what really matters.

Investment team

Programmatic sourcing has always been part of our DNA, with our internal data team surfacing the companies and entrepreneurs that align with our thesis. AI has simply supercharged that. Instead of spending three weeks “getting smart” on a sector, we can compress that into days and spend our energy pressure-testing assumptions with founders and customers, trying to solve and win over the highest conviction people.

Fund operations

Internally, we’ve built task forces that treat automation as core infrastructure. Portfolio reviews, reporting, and workflow prep are continuing to be rebuilt to run programmatically. The goal isn’t efficiency for its own sake. Rather, to free partners from reporting and give them more time for judgment and action.

Portfolio support

The biggest unlock has been in go-to-market diagnostics. By layering AI on top of our proprietary GTM data frameworks, we are now detecting early warning signs, conversion leaks, cohort behaviors, GTM team performance, often before the exec team itself. That lets us intervene in real time instead of in post-mortems. As we continue to collect more data, we’re building a proprietary portfolio data flywheel that compounds into sharper diligence, stronger sourcing signals, deeper insights, and more proactive portfolio support.

AI helps us prepare faster, see patterns earlier, and build compounding knowledge across the portfolio. But it is yet to tell us who has grit, ambition, or the moral compass to build a defining company. Those remain human calls, but ones we can spend more time thinking and debating about.

Have you already had any significant exits from AI companies? If so, what were the key characteristics of those companies?

We haven’t had a full exit yet from an AI-native company, though we’ve seen partial liquidity in several. What’s striking is that the companies compounding most effectively share a common trait beyond world-class teams: multi-layer product depth. They collect and control proprietary data, own the critical workflows, and build unit economics that actually improve with scale.

From our portfolio, VAST Data is a good example. It isn’t an “AI company” in the narrow sense, but it has become indispensable infrastructure. It's the foundation on which AI workloads run. Coralogix shows the same in observability: by embedding AI directly into developer workflows, it evolves into the system of record, not just another tool. Exodigo takes AI into the physical world, mapping the subsurface at scale in ways no one else can. Eleos Health does it in healthcare, generating privileged clinical data that rewires how providers work. DustPhotonics pushes on the hardware frontier: as AI workloads explode, bandwidth becomes the bottleneck, and their silicon photonics are critical to moving data across hyperscale data centers. Torq reimagines security automation by putting AI into the heart of workflows, allowing teams to orchestrate complex responses at machine speed.

Different markets, same pattern: generate proprietary data, own the workflow (or the choke point), and compound the business. That’s what we believe enduring AI companies look like.

Is identifying promising AI startups different from evaluating companies in your more traditional investment domains? If so, how does that difference manifest?

In some respects, evaluating AI startups isn’t much different - the best founders, in any domain, are obsessive about a customer problem and relentless about impact. They live inside the problem space and shape products around real needs. That hasn’t changed.

What is different in AI is the fragility of most businesses. A clever demo is not a company. If a product breaks the moment models get cheaper, faster, or open-sourced, it was never durable in the first place. The pace of innovation is ruthless. Whole categories can be rewritten in months, not years. You can’t just ask “is this good?” You have to ask “is this durable in the face of exponential change?”

In general, we believe that the real test is depth: do they control a proprietary data asset competitors can’t touch, do they own a workflow customers can’t rip out, and do their economics improve as they scale? The signal we look for is founder obsession paired with structural durability, and the rare ability to stay ahead when the ground itself keeps shifting.

What specific financial performance indicators (KPIs) do you examine when assessing a potential AI company? Are there any AI-specific metrics you consider particularly important?

The obvious numbers, revenue growth, retention, pipeline generation are strong indicators but often lagging. They tell you what happened yesterday. The real challenge in AI is whether the company has future-proof economics.

Gross margin is the first test. It shows you whether this is a software business or just GPU arbitrage. The second test is time-to-value: how quickly does a new customer see a measurable outcome? If it takes months, adoption dies; if it takes days, it compounds.

The truth is that the best KPI often is not financial at all, it’s distribution lock-in. If you own a proprietary channel, a dataset no one else can access, or a community that compounds, the financials will usually follow.

How do you approach the valuation of early-stage AI startups, which often lack significant revenues but possess strong technological potential?

We think of valuation as buying an out-of-the-money option. It often looks expensive today because you’re underwriting against a great team with a stellar product, albeit, with small revenues. But the bet is that, if the company wins, the payoff is asymmetric, it becomes an outlier that compounds far beyond what early numbers suggest. In the world of AI, this has become more expensive but for good reason.

As an example, take call centers. Traditional call center software sells licenses to make agents a bit more efficient. The TAM is capped by IT budgets. An AI-native call center isn’t a tool, it’s a labor substitute. The target market shifts from “software spend” to the global wages of human support reps. That’s trillions of dollars in wages and workflows. The TAM isn’t incrementally larger, it’s an order of magnitude larger and so the ability to build a big company grows accordingly.

That said, it’s important to ground ourselves in durability, e.g. gross margins net of inference costs, resilience as models evolve, but we’re comfortable paying up when the option is to own a market that only looks small until AI rewrites the economics.

What financial risks do you associate with investing in AI companies, beyond the usual technological risks?

One of the biggest financial risks in AI is mistaking speed for durability. Many companies will grow quickly, but very few will compound. If a business model weakens every time models get cheaper or competitors repackage the same foundation, it won’t last. The real signal is whether scale improves economics, whether data, workflows, or infrastructure compound into something competitors can’t easily copy.

Do you focus on particular subdomains within AI?

All the varying modality breakthroughs are brilliant, be it LLMs, voice, vision, or physical AI, but we don’t back “a modality.” We back companies applying them in ways that create real structural advantages for customers, which typically involves a combination of these breakthroughs working together.

The bigger idea is this: defensible companies aren’t built on the latest model, they’re built on solving problems the world actually needs solved. Our portfolio shows the pattern. Eleos Health began with voice and NLP to understand clinical conversations and is now layering additional modalities to rewire how providers deliver care. BigPanda started with machine learning for event correlation, added NLP for triage, and is now building a foundation model powered by autonomous agents to replace the IT analyst role altogether. Torq illustrates the same arc in security: from rule-based automation, to ML-driven detection, to natural language orchestration—and now toward fully autonomous SOC agents. In every case, the moat doesn’t come from a single breakthrough, but from compounding each wave of AI into deeper workflow ownership.

Beyond that, we’re watching sectors that have stubbornly resisted digitization (e.g. manufacturing, construction, logistics). Traditional software struggled to crack the ROI. What changes with AI isn’t just new modalities, it’s their convergence. Reasoning foundation models combined with computer vision can give machines the ability to both see and decide. Voice layered on top makes those systems usable in frontline environments where software has always failed. That fusion, reasoning plus perception plus interface, may finally flip the economics in industries where adoption once looked impossible, allowing them to leapfrog straight into AI-native systems.

How do you view AI’s impact on traditional industries? Are there specific AI technologies you believe will be especially transformative in certain sectors?

We briefly touched on it above, but we believe that some of the biggest AI opportunities won’t come from tech-first sectors, they’ll come from the industries software failed to transform over the last two decades. Healthcare, logistics, manufacturing, defence, and education are some examples. These markets resisted digitization because the economics never worked. AI changes that by collapsing the cost of repetitive, rules-driven work and unlocking workflows that once required armies of people.

But the bigger bet is on AI colliding with the physical world. Robotics and “physical AI” turn factories, warehouses, and even construction sites into adaptive systems. Our portfolio company Exodigo is a good example, using AI and advanced physics to generate subsurface maps at scale, solving a problem that had resisted digitization for decades. Other areas we are excited about and have been spending time in are biotech, where AI compresses the cycle time of discovery itself; what used to take years of trial-and-error in a lab can now be simulated and validated in silico. And in defense, autonomy at the edge changes the balance of power, systems that can sense, decide, and act faster than humans create entirely new capabilities. The enabling technologies are practical: retrieval over proprietary data, edge inference for real-time decisions, agentic systems with audit trails, multimodal models that mirror human input.

That said, the paradox is that AI won’t transform the jobs we always assumed were vulnerable, it will transform the ones we thought were safe. Lawyers, teachers, accountants, consultants: software never touched them because they were too human, too judgment-driven. But voice, vision, and reasoning combined don’t automate tasks, they replicate judgment. That’s the real disruption - AI is coming for the professions that defined the last century. A big blue ocean to build.

What specific AI trends in Israel do you see as having strong exit potential in the next five years? Are there niches where you believe Israeli startups particularly excel?

Israel is in a founder moment unlike anywhere else. On one side, repeat builders from Wiz, IronSource, Mobileye, Mellanox, people who’ve already built billion-dollar outcomes and are now chasing even bigger ones. On the other, a new generation shaped by the war, building with urgency and no-regrets execution. That mix of scar tissue and ambition is rare, and it’s the fuel for outsized companies.

If history proves anything, it’s that Israel doesn’t win in just one lane: semis, cyber, SaaS, consumer. The next wave of winners will likely continue to come from a multitude of fronts. We remain focused on infrastructure, where Israeli founders excel at solving hard technical bottlenecks, semiconductors, where the focus is now aimed at the compute limits of AI, cybersecurity, where military-trained engineers are uniquely positioned to secure the largest new attack surface since the internet; and defense, where necessity drives invention and ideas move from lab to battlefield faster than anywhere else.

That said, the surprise may be at the application layer: we’re already seeing ruthless execution in vertical software and even consumer AI, where Israeli teams are taking the same discipline they applied to enterprise and pointing it at global consumer markets.

Are there gaps or missing segments in the Israeli AI landscape that you’ve identified? What types of AI founders are you especially looking to back right now in Israel?

We are in the early phases of AI-native application companies coming out of Israel. We are excited to see what the future holds. As for founders, we are looking to partner with teams who are humble, hungry, and smart…in that order.