The Innovator's Dilemma in the AI Era: Why Your Best Customers Will Kill Your Next Product

Clayton Christensen published The Innovator’s Dilemma in 1997. His central claim was so counter-intuitive it took decades to become mainstream: successful companies don’t get disrupted because they’re badly managed. They get disrupted because they’re brilliantly managed — for their existing customers.

That’s the dilemma. The same discipline that makes you successful in your current market — listening to your best customers, investing in what drives highest returns, optimising margins — is what prevents you from investing in the cheap, ugly, low-margin, low-end product that will eventually eat your lunch.

In 1997 this took a decade to play out. In 2026, with AI collapsing the cost of building a credible software product to near zero, it can happen in eighteen months.

The Innovator’s Dilemma is Clayton Christensen’s 1997 theory that successful companies get disrupted not because they are badly managed, but because they are brilliantly managed — for their existing customers. The same discipline that makes incumbents succeed at sustaining innovation (listening to best customers, prioritising highest-margin investments) systematically prevents them from investing in the disruptive low-end products that eventually take their market.

Sign up (free) to protect capacity for disruptive bets

What the Innovator’s Dilemma Actually Says

Christensen’s thesis rests on a distinction between two kinds of innovation:

• Sustaining innovation — improving an existing product along the dimensions existing customers already value. Faster chips. More features. Higher resolution. This is what mature companies are excellent at.
• Disruptive innovation — a new product that is worse than the incumbent on the dimensions mainstream customers care about, but better on a different dimension (price, simplicity, accessibility) that a non-consuming or underserved segment values. Over time, the disruptive product improves enough to satisfy the mainstream — and the incumbent wakes up to find they’ve lost the market.

The foundational case study is the hard disk drive industry — Christensen’s own doctoral research. Each generation of smaller drives (14" → 8" → 5.25" → 3.5" → 2.5") was initially too small, too low-capacity, and too low-margin for incumbent customers. Incumbents systematically dismissed the smaller format. New entrants captured the new application, improved rapidly along capacity, and destroyed the incumbent. Seagate, IBM, and Control Data each lost their leading positions this way — not because they were badly run, but because listening to their best customers led them straight off the cliff.

Other canonical examples: Nucor’s mini-mills undercut integrated steel producers by entering at the low end (rebar) and moving up market over decades. Toyota entered the US not with the Lexus but with the Corolla — dismissed as a cheap, unambitious car that couldn’t threaten American auto-makers. Minicomputers disrupted mainframes, PCs disrupted minicomputers, smartphones disrupted PCs. Closer to our time: Kodak invented the digital camera and couldn’t disrupt its own film business; Blockbuster ignored Netflix; Nokia and BlackBerry watched the iPhone eat them alive.

Low-End, New-Market, and Market-Creating Disruption

Christensen later refined the theory into three sub-types, which the Christensen Institute continues to teach:

• Low-end disruption — the disruptor targets overshot customers at the bottom of the existing market (Nucor in steel, Walmart vs. department stores)
• New-market disruption — the disruptor targets non-consumers who previously lacked access (personal computers vs. timeshare mainframes; cheap mobile phones in emerging markets)
• Market-creating disruption — an overlap with new-market; creates an entirely new consumption class (the Model T automobile; streaming video)

The three-way split matters because the defensive response differs. Low-end disruption requires you to defend margin at the bottom or cede the bottom deliberately. New-market disruption requires you to enter the new market yourself before the disruptor graduates up to your core.

The Jobs-to-Be-Done Connection

Christensen later argued that disruption is best understood through Jobs to Be Done — what job is the customer hiring the product to do? Incumbents optimise for the jobs their best customers currently hire them for. Disruptors identify jobs that are either unmet or poorly served and build products that do those specific jobs better. A disruptive product rarely wins on absolute performance; it wins on job-fit for an underserved population. For product teams, JTBD is the diagnostic tool: if you can’t name the specific jobs your non-consumers need done, you won’t see the disruption coming.

Christensen’s Own 2015 Caveat

In a 2015 Harvard Business Review retrospective, Christensen himself clarified that not every competitive disruption is “Innovator’s Dilemma” disruption. Uber, he argued, is NOT disruption in his sense — it entered at the high end of the taxi market, not the low end, and attacks incumbents head-on rather than from below. This matters: the dilemma is specifically about being dethroned by an initially-inferior entrant that you rationally chose to ignore. Generic “new competitor” stories don’t qualify.

The dilemma is that rational management makes self-disruption almost impossible:

• Your best customers don’t want the disruptive product (by definition — it’s worse on the dimensions they value)
• Margins on the disruptive product are lower than on your core product
• Your sales force doesn’t want to sell it
• Your finance team can’t justify the investment on a returns basis
• Every internal decision-making process optimises for your core product

So the disruptive bet — which you logically know you should fund — never gets funded. Because everyone is doing their job well.

Why AI Makes the Dilemma Worse, Not Better

Here’s the 2026 reframe that every article you’ll read on the Innovator’s Dilemma completely misses.

Christensen’s model assumes the disruptor has a cost and time disadvantage on the build side. In 1997, building a credible low-end alternative to an enterprise software product required funding, engineers, and years. By the time the disruptor ramped up, the incumbent had (in theory) time to respond.

AI has destroyed that assumption. An AI-native disruptor in 2026 can:

• Build a working version of your core product in weeks, not years
• Launch with 10% of your feature set at 20% of your price
• Attack a low-end segment you’ve implicitly abandoned
• Iterate faster than you can respond, because they have no legacy code, no enterprise customers holding the roadmap hostage, and no organisational inertia

The build side of Christensen’s dilemma has collapsed. What hasn’t collapsed is the sell side — distribution, trust, brand, reference customers. Which means disruptors still have to do the hard work of GTM, but the economics of attempting it are now an order of magnitude cheaper.

The practical consequence: the number of credible disruptive threats facing every incumbent has exploded. What used to be a theoretical risk to plan for over 5–10 years is now a real risk to plan for over 18–24 months. The same economics that make the chasm wider for disruptors also make them more numerous.

And here’s the thing most boards miss: AI doesn’t just empower disruptors. It empowers you, too — if you have the discipline to use it.

The Roadmap Response: Protect Your Own Disruption

If you accept Christensen’s thesis, the roadmap question becomes:

What are our disruption bets, and are we protecting them from our own customers?

This is where the Three Horizons framework earns its keep. Horizon 1 (defend the core) will always have the loudest voices — they are your current customers, your highest-revenue account managers, and your most senior product leaders. H1 will always outshout H3 (speculative disruption bets). That’s not a bug; it’s the dilemma operating exactly as Christensen described.

The only way to protect H3 investment is structural:

• A dedicated team on the H3 product, not a 20%-time project. See the minimum viable new-product team — 2 engineers and a PM, ring-fenced.
• A separate budget line that doesn’t compete for quarterly prioritisation with H1 revenue features.
• Different success metrics — reference customers, time-to-first-dollar, not revenue contribution. Applying H1 metrics to an H3 product will kill it before it has a chance. The validation-stage metrics are specifically covered in the early-stage validation cluster — riskiest assumption tests and assumption mapping are the tools that produce the evidence an H3 team should be measured on.
• A separate operating model — see below.
• Protection from customer-led priority whiplash . The moment your best customer’s urgent request re-prioritises the H3 team, the dilemma has won.

Note that this is exactly what Christensen proposed in The Innovator’s Solution (his follow-up): spin out the disruptive bet into a separate organisation with its own P&L, its own metrics, and its own leadership. Most companies skip this and then wonder why the internal disruption bet kept getting deprioritised. The answer is: because your governance worked exactly as designed.

The Operating Model Mismatch

This is where Marty Cagan’s work on the product operating model connects directly to Christensen.

A sustaining-innovation team and a disruptive-innovation team need fundamentally different operating models:

Most companies apply their H1 operating model to H3 bets. The result: the disruptive product gets measured on revenue, fails those metrics (because it’s a year from revenue), and gets killed. The dilemma wins again.

Attack Your Own Adjacent Pools of Value (Before a Competitor Does)

The PE-era reframe of the Innovator’s Dilemma is offensive, not defensive:

Given the portfolio of assets we own — customer relationships, distribution channels, data, brand — what adjacent pools of value can we attack with new products, before a competitor attacks our core?

This flips Christensen’s framework into an opportunity lens. Your distribution asset is exactly the thing AI-era entrants don’t have. If you can move fast into adjacent value pools — new segments, new products, new use cases — you deny the disruptor the easy low-end entry they’d otherwise use. Your Ansoff Matrix market-development and product-development quadrants become the strategic map of your counter-attack.

The companies I admire most in PE work are the ones that proactively cannibalise their own products before a competitor does. They allocate capacity to disruption of themselves, before the market forces the issue. This is uncomfortable for the H1 team but almost always cheaper than defending against external disruption after the fact.

Try RoadmapOne to protect Horizon 3 capacity — free sign-up

The Board / NED Diagnostic for the Dilemma

When I look at a portfolio company’s roadmap from a board seat, here are the questions I ask to test whether the company is at risk of being disrupted:

1. What percentage of our capacity is on H3 (disruptive) bets? If the answer is zero or “hard to say”, that’s the first red flag.
2. Do we have a dedicated team on H3, or is it side-of-desk? Side-of-desk H3 is not H3; it’s theatre.
3. Who are the credible entrants in our market? If the answer is “nobody, we don’t have real competition,” the board should be actively suspicious. Every mature market has AI-native entrants in 2026.
4. What would it cost for a competitor to build our MVP? If the answer is “months, not years,” the moat isn’t the product — it’s distribution, trust, and scale. Those are the assets to invest behind.
5. How much of our roadmap is driven by our top 10 customers’ feature requests? If it’s >60%, the dilemma has already taken over. Your best customers are locking you into the sustaining-innovation path.
6. Can we name the adjacencies we’re attacking? Not vaguely — by name, with dedicated teams and targets. If the answer is vague, there’s nothing to audit and the dilemma will win by default.

A healthy answer looks like: 15–20% of capacity on H3, with 2–3 named disruptive bets each with a dedicated minimum viable team, insulated from quarterly H1 pressure. Anything less and you’re betting that the status quo will hold — which, in 2026, is a bet against your own survival.

Common Failure Modes

Even companies that understand the dilemma trip over these:

• “We already innovate” theatre. Re-badging H1 feature work as “innovation” on a slide. The test: is there a dedicated team with a different operating model and different success metrics? If not, it’s theatre.
• Internal innovation labs disconnected from the roadmap. Labs that don’t have a path to becoming a real product. They generate demos and press releases; they don’t generate new revenue streams.
• Over-indexing on patents or IP. IP is a legal moat, not a strategic one. Most disruption in software doesn’t rely on beating you on IP — it beats you on price, simplicity, and distribution.
• Refusing to cannibalise. Deciding not to launch a cheaper / simpler version of your own product because it “would hurt our core business”. Fine — but a competitor will launch it instead, and hurt your core business more.
• Putting your best people on H1. The rational decision — the best people drive the most revenue — is also the decision that guarantees the dilemma wins. You need your best people on H3 at least some of the time.
• Feature factory behaviour. A roadmap dominated by prospect-specific features from your core customers is exactly how the dilemma manifests operationally. If that’s you, the board should intervene.

How RoadmapOne Helps

RoadmapOne was built to make the dilemma visible. Tag objectives by Three Horizons or Run/Grow/Transform and the analytics tell you exactly what percentage of your capacity is on disruptive bets vs. sustaining the core. Most companies are surprised — their stated strategy is 20% on Transform; the capacity grid shows 3%. That gap is the dilemma operating in real time.

Frequently Asked Questions

What is the Innovator’s Dilemma in simple terms?

Successful companies get disrupted by new entrants because doing the sensible thing — investing in what their best customers want — means they systematically under-invest in the new, cheap, simple products that eventually take over their market.

What are some examples of the Innovator’s Dilemma?

Kodak and digital photography (Kodak invented the digital camera but couldn’t disrupt their own film business). Blockbuster and Netflix. Nokia and the iPhone. Blackberry and touchscreens. In each case the incumbent saw the threat and was unable to respond at scale — not because they were dumb, but because their governance structurally prevented disrupting their own core.

How is the Innovator’s Dilemma different from disruptive innovation?

The Innovator’s Dilemma is the predicament faced by incumbents. Disruptive innovation is the mechanism that creates the predicament. The dilemma is the organisational challenge; disruption is the market phenomenon causing it.

How do companies solve the Innovator’s Dilemma?

Usually by spinning out the disruptive bet into a separate organisation with its own budget, metrics, and operating model — as Christensen recommended in The Innovator’s Solution. Internal attempts almost always fail because the core organisation’s governance crowds out the disruptive bet. See also: dedicated H3 teams with protected capacity and outcome-based objectives .

Does AI make the Innovator’s Dilemma more or less of a problem?

Dramatically more. AI has collapsed the cost of building a disruptive alternative to near zero, which means the number of credible disruptive threats facing any incumbent has multiplied. The dilemma hasn’t changed — but the timescale has compressed from decades to months.

Is the Innovator’s Dilemma still relevant today?

More than ever. If anything Christensen’s original work understated the problem. The rise of AI-native startups, platform business models, and low-cost global distribution has made the conditions for disruption more common, not less. Any incumbent not actively protecting disruptive bets is betting on the status quo.

Sign up (free) to tag your roadmap by Horizon and expose the dilemma

Conclusion

The Innovator’s Dilemma is not a strategy problem. It is an allocation problem. Everyone in your company can understand the theory. Almost nobody builds governance that protects disruptive bets from the gravitational pull of sustaining-innovation work. In 2026, with AI accelerating the timescale on which disruption plays out, that allocation discipline is the single biggest determinant of whether your company survives the next decade.

The good news: the fix isn’t expensive. A dedicated minimum viable team on two or three H3 bets, with their own metrics, their own operating model, and protection from quarterly pressure, costs a rounding error relative to the existential threat of being disrupted. The bad news: most companies still won’t do it. That’s why the dilemma keeps claiming new victims.