Build the cannibal before it eats you

Transformation is the comfort blanket boards reach for when disruption arrives.

It feels responsible. It sounds strategic. It lets everyone keep their jobs while promising better versions of themselves in three to five years. The problem is that 88% of transformations fail to achieve their original ambitions, according to Bain's 2024 research. And the 12% that succeed? They take so long that the technology they're implementing is already obsolete by the time it ships.

There's another path. One that most boards dismiss as too risky until disruption forces their hand. Building an AI-native venture that cannibalises your own business before someone else does.

The evidence is clear. Corporate venture building achieves a 66% success rate when executed by expert builders, according to BCG Digital Ventures. New ventures generate disproportionate value, contributing 16% of enterprise value while representing just 13% of revenue. And the largest new ventures built by established companies in the past decade have achieved 1.5x the revenue of the largest startups launched in the same period.

Better a cannibal you control than one that lives outside your walls, eating your lunch.

The transformation failure thesis is no longer debatable

The consulting industry has reached remarkable consensus. McKinsey reports 70% failure rates across multiple studies spanning 2018-2024. BCG found only 30% success in their analysis of 850+ companies. RAND Corporation's 2024 research on AI projects specifically found 80%+ failure rates, twice the rate of non-AI IT projects.

The root causes cluster around organisational rather than technical failures. BCG's research reveals what they call the "10-20-70 Principle": AI success requires only 10% algorithms and 20% data and technology. The remaining 70%? People, processes and cultural transformation. That's the bit that breaks.

Time-to-completion data reveals a structural mismatch. Enterprise-wide transformations typically require five-plus years, with some extending to seven to thirteen years in complex programmes. Only 12% of organisations sustain transformation goals for more than three years. Average CEO tenure sits at 4.8 years for S&P 500 companies. Do the arithmetic. Digital transformation takes two to three times longer than the executive who sponsors it will be in post.

I've seen this pattern repeatedly. The new CEO arrives with transformation ambitions. Year one is strategy and planning. Year two begins implementation. Year three hits organisational resistance. Year four the CEO moves on. Year five the programme quietly dies or gets relabelled.

Martec's Law guarantees your transformation will be obsolete

Scott Brinker articulated something in 2013 that every board should have tattooed on their foreheads: technology changes exponentially while organisations change logarithmically. The gap between what technology can do and what organisations actually do widens continuously.

Think about what this means for a five-year transformation programme. The AI capabilities available when you finish will be five generations ahead of what you designed for. The competitors who started building AI-native from scratch will have operated with that newer technology from day one.

Technical debt makes it worse. McKinsey research found that tech debt amounts to 20-40% of entire technology estates before depreciation and accounts for approximately 40% of IT balance sheets. Engineers at some companies spend up to 75% of their time paying the tech debt "tax". That's three-quarters of your technical workforce maintaining the past rather than building the future.

When tech debt exceeds 50% of tech asset value, McKinsey recommends considering a "greenfield stack" approach. Starting fresh rather than attempting incremental remediation. This aligns with Brinker's 2020 amendment to his law: the concept of "punctuated equilibrium reset", where organisations make evolutionary jumps only when cataclysmic events force dramatic adaptation.

COVID-19 demonstrated this. Organisations accomplished in weeks what normally takes years. The forcing function wasn't a better strategy deck. It was existential necessity.

For many health insurers, an honest assessment of their technology estate would reveal they've passed the reset threshold. It's time to stop pretending the house can be renovated when the foundations need replacing.

Building new beats transforming old

McKinsey's 2024 Global Survey on Corporate Venture Building found that expert builders achieve twice the success rate of novices and generate 12x more revenue in a venture's fifth year despite investing only 2x the capital before breaking even.

The value creation differential is stark. Companies investing 20% of growth capital into building ventures achieve 2 percentage points higher revenue growth, rising to 2.5 percentage points for billion-pound organisations. Only 38% of companies invest at this level. That's a lot of unrealised potential sitting on balance sheets.

Serial building creates compounding advantage. Organisations building three-plus ventures per year achieve a 2.8:1 success ratio of successful to underperforming ventures. Expert builders are 3x more likely to have established frameworks, 2.6x more likely to have dedicated ring-fenced funding and 2x more likely to grant decision-making independence.

Scott Anthony's "Dual Transformation" framework from Innosight provides the theoretical foundation: Transformation A repositions today's business for resilience while Transformation B creates a new growth engine. They're linked by a Capabilities Link that leverages difficult-to-replicate assets.

The insight from McKinsey's research validates this: new builds supported by parent companies combine incumbent assets with startup agility. You get the best of both worlds. The domain expertise and customer relationships of the incumbent. The architectural freedom and speed of the startup.

The demo trap is where AI initiatives go to die

MIT's NANDA Study from July 2025 found that 95% of GenAI pilots fail to deliver measurable P&L impact, based on 150 interviews with leaders, a survey of 350 employees and analysis of 300 public AI deployments. Gartner predicts that 30% of GenAI projects will be abandoned after proof of concept by end of 2025.

I call it the demo trap. The pattern is painfully predictable. Innovation team identifies opportunity. Pilot gets funding. Pilot succeeds in controlled conditions. Then nothing. The pilot never scales because scaling would require changing how the organisation actually works. And nobody signed up for that.

Time-to-scale differentiates leaders from laggards. AI leaders move from idea to scale in 5-7 months versus 15-17 months for laggards, according to BCG's 2023 research. Large enterprises experience 9+ month pilot-to-scale cycles compared to approximately 90 days for mid-market firms.

Here's the counterintuitive finding. MIT NANDA's study found that internal AI builds achieve only 33% success rate while vendor partnerships and purchased solutions achieve 67%. Building AI-native capabilities through ventures with external partnerships may outperform internal transformation efforts. The "build it ourselves" instinct that most enterprises default to may be precisely wrong.

Use cases bolt-on. They don't transform. You can deploy AI for claims automation, fraud detection, customer service. These are worthwhile. But they're optimisation, not reinvention. Faster rubbish is still rubbish. The opportunity is to reimagine the entire enterprise: workflows, people, data, architecture. Not just sprinkle AI on top of processes designed for humans pushing paper.

The agentic organisation demands flow, not functions

McKinsey's September 2025 research on "The Agentic Organization" identifies a structural transformation from functional silos to outcome-aligned agentic teams. The traditional model of organisations built around functional silos with cross-functional teams constrained by handovers is giving way to something fundamentally different.

In the agentic organisation, a human team of 2-5 people can supervise an agent factory of 50-100 specialised agents. Organisation charts built around hierarchical delegation pivot toward agentic networks, work charts based on exchanging tasks and outcomes rather than managing people up reporting lines.

This is not automation as usual bolted on top of existing processes. It's a redesign of end-to-end processes with humans "above the loop" for strategic oversight. The potential is to bring marginal cost toward the cost of compute. Not incrementally cheaper. Fundamentally different economics.

McKinsey found that 89% of organisations still operate with industrial-era hierarchies. 9% have adopted digital-era agile models. Only 1% operate as decentralised networks, the organisational structure most compatible with agentic operations.

I think of this as "flow architecture". Functions and processes are artificial silos built for a world driven by people and manual processes. In an AI-native enterprise, you need continuous adaptation in shorter and shorter learning loops. We're entering the era of programmable intelligence. If architected correctly, domain matters less. An AI-native infrastructure can address new verticals and industry segments for growth in ways that functional organisations simply cannot.

You cannot retrofit flow onto functional architecture. You have to build for it from the start.

Simulate the future before committing capital

Here's an underappreciated opportunity. Incumbent enterprises sit on decades of operational, financial and customer data. This data can be used to simulate and rehearse how an AI-native enterprise could work before releasing it into the wild.

McKinsey reports that 70% of C-suite technology executives are exploring digital twin investments, with potential to increase decision-making speed by 90%. TCS's TwinX platform enables organisations to test business decision outcomes in virtual A/B testing mode and conduct what-if and if-what analyses before committing resources.

Think about what this means for a health insurer considering an AI-native venture. You could model how AI-first underwriting would perform against your historical book. You could simulate how agentic claims processing would handle the edge cases that consume human adjuster time. You could test pricing strategies and coverage designs against actual customer behaviour.

This simulation capability is an accelerant. It reduces uncertainty before market release. Every function needs rethinking, but you can rehearse that rethinking before betting the company on it.

The enterprises that master simulation will make fewer expensive mistakes. They'll know which AI-native designs work before competitors figure it out through trial and error in market.

Health insurers are moving in the wrong direction

Menlo Ventures' October 2025 survey of 700+ healthcare executives reveals that health insurers are uniquely unprepared for what's coming. Health systems shortened AI buying cycles by 18% from 8.0 to 6.6 months. Payers lengthened cycles by 20% from 9.4 to 11.3 months.

Payer AI adoption stands at only 14% versus 27% for health systems. Payers represent just $50 million (5%) of the $1.4 billion healthcare AI spend compared to $1 billion (75%) for providers.

The opportunity cost is staggering. McKinsey's payer-specific projections suggest that for every £10 billion of payer revenue, AI could deliver £150-300 million in administrative cost savings, £380-970 million in medical cost savings and £260 million to £1.24 billion in revenue increase. Net administrative costs could fall 13-25% while medical costs could decline 5-11%.

Yet health insurers move slower than everyone else. The sector that faces the most existential threat from AI-native competitors is responding with the least urgency.

InsurTech "build new" strategies have largely failed in health insurance, but for instructive reasons. Bright Health Group's June 2021 IPO valued the company at $11 billion. Net losses of $1.4 billion in 2022 led to market cap collapse to $52 million by December 2023. Oscar Health lost $610 million in 2022. Clover Health lost $339 million.

The failure wasn't the "build new" strategy. It was startups who didn't understand they were in a risk management industry. They lacked scale for provider negotiations, risk adjustment capabilities and understanding of structural market dynamics. This is precisely why incumbent-backed ventures have an advantage. They inherit the domain expertise that pure startups have to learn through expensive failure.

The controlled cannibal beats waiting to be eaten

Netflix transformed from DVD-by-mail to streaming and delivered 1,567% revenue growth from $1.2 billion in 2007 to $20 billion+ in 2019. Tencent set up a new team in Guangdong specifically to cannibalise their existing QQ product, creating WeChat with 690 million+ active users. Apple willingly cannibalised the iPod with the iPhone and became the world's most valuable consumer company.

Companies that failed to self-cannibalise became case studies in disruption. Kodak invented the digital camera in 1975 but held 80% market share in U.S. film and feared cannibalisation, declaring digital photography "the enemy". Nokia held 50%+ global phone market share in 2007 but dismissed a touchscreen prototype. One year later Apple launched iPhone. Blockbuster rejected a $50 million acquisition offer from Netflix.

However, HBR research by Wharton professors found that self-disruption can actually hurt companies with high stocks of key assets devoted to traditional business and those in highly competitive environments. The implication: timing and context matter more than categorical prescriptions.

Corporate spin-out structures dramatically improve outcomes. Research found that corporate spin-outs achieve 36% successful exit rate versus 21% for peers, with time to exit at 4.9 years versus 9.7 years. They reach first equity rounds in under a year versus 2.6 years and raise significantly more capital in the same timeframe.

Board-level sponsorship proves essential. Expert venture builders are 2x more likely to grant ventures decision-making independence and 1.4x more likely to have C-suite champions. The ventures that succeed have air cover from the top. The ones that fail get suffocated by corporate antibodies.

Innovate from strength, not distress

The evidence strongly supports investing in transformation while the core business remains healthy. McKinsey research on crisis innovation found that companies maintaining innovation focus through the 2009 financial crisis outperformed market average by 30% over subsequent 3-5 years. Top economic performers are 63% more likely to innovate at scale by building or acquiring new businesses outside current industries.

The central paradox: successful companies have resources but lack urgency, while declining companies have urgency but lack resources. Kodak, Nokia and Blockbuster had the resources to transform when they held dominant market positions. They didn't have the urgency. By the time urgency arrived, resources had evaporated.

PwC research quantifies this: only 58% of companies believe they modified their business model at the right time; others moved either too early or too late. Top-third performers achieved 71 percentage points of performance premium versus industry peers.

The optimal time to build an AI-native venture is when you're still profitable but seeing early warning signals. You have capital to fund the venture properly. You have talent you can redeploy. You have customer relationships to leverage. You have domain expertise to transfer. And you have time to iterate before the market forces your hand.

If you wait until you're under pressure, it's too late. You'll be trying to fund innovation from declining cash flows. You'll be trying to attract talent to a sinking ship. You'll be trying to convince customers to trust you with their future when they can see your past crumbling.

Don't wait. Build now. Build from strength.

The honest assessment before you build

The path you take depends on brutal honesty about four dimensions:

Leadership capability. Do you have executives who can operate with startup uncertainty while leveraging corporate assets? Most corporate leaders are selected for their ability to manage complexity and minimise risk. Venture building requires comfort with ambiguity and appetite for calculated bets. If you don't have this leadership DNA, you need to acquire it or acknowledge that transformation, despite its lower success rate, may be your only realistic option.

Capital availability. Can you ring-fence 8-12 years of patient capital without the venture being raided during quarterly earnings pressure? McKinsey's research shows expert builders are 2.6x more likely to have dedicated ring-fenced funding. If your capital allocation process treats ventures as discretionary spending that gets cut when the core underperforms, don't start.

Talent and workforce readiness. Do you have people who can work in AI-native environments, or will you need to acquire them all externally? Hybrid teams that combine domain expertise from the parent with AI fluency from the market tend to outperform homogeneous teams from either source. McKinsey projects 75% of current jobs will require redesign, upskilling or redeployment by 2030.

Data and tech readiness. Is your data an asset or a liability? Can it be extracted from legacy systems and made available to AI-native architectures? If your tech debt exceeds 50% of tech asset value, McKinsey's greenfield recommendation applies. Starting fresh may be faster than excavating data from systems designed before anyone imagined current AI capabilities.

If you score poorly on all four dimensions, transformation may be your only path, but go in with eyes open about the 88% failure rate. If you score well on at least three, building an AI-native venture deserves serious consideration as your primary strategy rather than an innovation side project.

The choice facing every health insurance board

You have three strategic options. Transform the core. Build an AI-native venture. Do both through dual transformation.

The aggregate evidence supports AI-native venture building as the superior option when executed by expert builders with board sponsorship and patient capital. Transformation failure rates of 70-88% contrast with venture building success rates of 66%. New ventures generate disproportionate value (16% of enterprise value from 13% of revenue) while incumbent-built ventures achieve 1.5x the revenue of standalone startups. Martec's Law and tech debt data show that transformation timelines guarantee technology obsolescence, while demo trap statistics reveal that 70-95% of AI initiatives fail to deliver measurable business value through traditional implementation.

For health insurers specifically, the sector's 20% longer buying cycles and 5% share of healthcare AI investment indicate a conservative approach that may prove strategically fatal. InsurTech failures demonstrate that pure startups lack the domain expertise and scale advantages that incumbents possess, but incumbent transformation programmes face the same 88% failure rate as other industries.

The synthesis points toward a middle path. Creating AI-native ventures that inherit incumbent domain expertise and assets while operating with startup governance, dedicated teams and architectural freedom.

Better to build the cannibal yourself than wait for someone else to build it.

The clock is running. Your competitors aren't waiting. The question for every health insurance board is not whether disruption is coming. It's whether you'll be the disruptor or the disrupted.

Choose accordingly.