.png)
Chegg built a $12 billion business helping students rent textbooks, complete homework, and find tutoring. Then suddenly, ChatGPT launched, and students could get all of it for free, blindsiding Chegg into collapse. In 2023, Chegg's CEO became one of the first executives to publicly blame AI for destroying his company and by 2026, the share price had fallen 99%.
Chegg had the context. It just never built anything that used it. Its system matched a question to an existing answer, the same one for every student who asked it, regardless of what it already knew about that student's specific struggles. Context only becomes a moat once a company builds the infrastructure to act on it.
Chegg will not be a one-off. Incumbents have always had time to respond to a threat, because building a company large enough to be one used to take years. Now, that time has disappeared.
Personio, one of Europe's fastest-growing software companies, took six years to reach $100M in revenue. Legora, in Cherry's portfolio, did it in 18 months. A challenger can now reach serious scale before an incumbent has finished deciding what to do about it.
But speed cuts both ways. A challenger can build fast precisely because it starts with nothing to lose and nothing to build on. An established company starts with the opposite: customers, proprietary data, and trust accumulated over years. These are things a fast-moving newcomer cannot replicate at any speed. Used well, and built into something that acts on them, those assets turn AI from a threat into a moat.
You can't out-ship AI natives
AI is attacking established software companies from several directions at once. Customers who once needed a product to find an answer now just ask ChatGPT. Agents handle the CRM updates, call summaries, and follow-up drafts that software used to do, with no product in the loop. And AI-native startups are rebuilding whole categories from scratch, with smaller teams and prices incumbents can't match.
The stickiness is going too.
Products that were hard to leave because analysts and consultants ran their work through them are losing their grip, now that AI can do that work directly, and buyers are increasingly choosing whatever feels AI-native, regardless of features or track record.
So how do established software companies win against AI-natives? The answer is context; everything an established software company has built up over years that an AI native is starting without.
Acting on context is the moat
Frontier models are converging and features are standardising fast. Build your defensibility on the model and you are building on sand.
Context is the exception. The proprietary data, customer relationships, and operational knowledge you have built over years cannot be copied at any speed, and the companies using it are already turning it into revenue.
A procurement company in Cherry's portfolio rebuilt its product as an agentic AI system on top of years of data on parts, prices, and supplier deals. That data took it from zero to €5M in AI revenue in six months.
A financial services company used the same advantage differently. At renewal, it moved its 4,000 customers onto an AI-powered version of its product, and those who upgraded now pay double. It then absorbed the compliance and reporting work those customers had been outsourcing, adding a second revenue line.
The mechanism is the same in both. The more an AI system learns about how a customer operates, the harder that customer is to lose.
Regulation is context in another form. An AI-native startup entering financial services, healthcare, or law spends years clearing compliance before it can sell anything. An established company already has. The cost that always looked like a burden is a barrier to entry that works in your favour.
Two Paths to Adding AI to Your Business
There are two ways to respond to AI natives.
Transform your core product
Hire one or two senior AI engineers into your product team. They go through every feature you have and ask what AI can do better, faster, or autonomously. Dash0 rebuilt their entire architecture on an open standard, then layered autonomous intelligence on top. Incumbents like Datadog can't follow without rebuilding from scratch and breaking everything their existing customers rely on.
The gap between AI-fluent teams and everyone else shows up directly in engineering output. Dash0's AI-native engineering team ships 35 pull requests per month per engineer. That's well above the top 10% of high-performing engineering orgs generally, who land around 22, and more than double the industry median of 12.4. Teams with daily AI users see roughly 10 more PRs per month than teams without them.
One Cherry portfolio company measured this in a pilot: the time to ship a code change fell 50%, from four days to two. Each change shipped was also bigger and more complete, nearly double the size of before, and came with 6 to 14 times more automated testing than the team's previous standard. The team produced 50% more output with the same headcount.
Launch a new AI product built on your proprietary data
Your contextual knowledge is your unfair advantage when building something new. Legora built a generative AI product for legal teams from scratch. They skipped the self-serve and gradual rollout approach, went directly to enterprise customers, and reached $100M ARR in 18 months.
Each path requires different teams, sales motions, and ways of measuring success. Pick one and commit.
The AI Maturity Framework
Most companies have licensed tools and a handful of power users who swear by ChatGPT. That is L1 of the AI maturity framework, a six-level scale that tells you how deeply AI is embedded in how your organisation actually operates. At L1, a few employees save time but it has no organisational impact.
The real value compounds at L3, when AI can see across the whole organisation, agents act across systems, and non-engineers start shipping their own workflows.
Cherry portfolio companies operating at L3 are doubling revenue without adding headcount. A Series B SaaS company moving from L2 to L3 is worth an estimated €1.2M per year in combined cost savings and additional revenue.
Industry-wide, AI adoption sits well below the L3 target in sales, marketing, customer success, revenue operations, and HR. Sales and SDR average 49% adoption, against an 80%+ target. Marketing is closer at 60%, still short of 85%+. Customer success trails at 50% versus 80%+. The bigger gaps are in revenue operations and finance (35% against 70%+) and people and HR (30% against 65%+).
The highest-value workflows are still being done manually. A rep has to initiate every CRM query by hand; nothing runs on its own. Campaign briefs start from a blank page rather than pulling from shared brand context. Support tickets sit in a queue until a person triages them. Finance updates its forecasts once a month instead of watching them move with the CRM in real time.
Cherry's portfolio data shows that closing the gap is worth roughly €540K a year in engineering and support cost savings, plus €675K in additional revenue from GTM efficiency gains, for a combined €1.2M annual impact. Portfolio companies that have made this move did it in about six months, from first pilot to scaled rollout.
The two paths described above are how you get there. But knowing which level you are starting from determines how much ground you have to cover to reach L3.
The Founder Mandate
Most companies hire a head of AI, make an announcement at an all-hands meeting and expect the transformation to happen. Every AI transformation Cherry has seen succeed has been driven by the founder through their daily work.
That means three things in practice. The founder uses AI visibly, screen-sharing their own workflows so the team can see how. AI capability becomes a filter in hiring, so candidates solve a role-relevant problem with AI in real time. And, the company measures whether any of it is working, with before-and-after numbers and quality guardrails.
One Cherry portfolio founder declared 'Agent Week'. Every employee, engineering and non-engineering, had to have hands-on experience with AI tools on real work by Friday. By the end of the week, people who had never opened an AI tool in their working lives were using one daily.
The Window is Narrowing
Non-engineers are shipping agents. Finance teams are closing months in two days instead of eight. The gap between them and the companies still at L1 is widening every quarter.
Chegg had years of context. By the standard most companies use to measure their own defensibility, it should have been unbeatable. It wasn't. Chegg knew a given student had gotten the same type of algebra problem wrong three times that semester. The answer it gave her was identical to the one it gave a student seeing the concept for the first time.
The companies that win this transition will hold the same kind of context Chegg had. Some will rebuild what they already sell. Others will sell something new on top of what they already know. Either way, the difference is they'll build the infrastructure to act on it.
The first step is to be honest about how far your company has come with AI. Everything else follows from that.
.png)
