For a decade the Rule of 40 has been the single number public-market investors use to judge a software business: revenue growth plus operating margin should clear 40%. A new Bain & Company analysis argues that benchmark is now breaking under AI, with rising inference costs eroding the high-margin side and slowing market expansion pulling at the growth side at the same time. For UK software boards — whether they sit in London-listed platforms, PE-backed scaleups, or bootstrapped vertical SaaS — the implication is uncomfortable. The financial model that justified valuations, hiring plans, and exit timelines is quietly being rewritten, and most management teams have not yet reset expectations with their investors.
The benchmark that defined a decade is becoming unreliable
The Rule of 40 worked because SaaS unit economics were unusually clean. Once a customer was acquired, serving them cost close to nothing, gross margins sat above 75%, and scale produced leverage almost mechanically. Investors could model trade-offs between growth and profit with confidence because the underlying cost structure was predictable.
Generative AI has broken that predictability. Every prompt, every agent action, and every model call introduces a variable cost that scales with usage rather than with price. Bain cites a marketing technology company where revenue grew 38% year-on-year whilst costs grew 349%, much of it AI infrastructure spend. That is not a rounding error. It is the inversion of the economic logic that made SaaS attractive in the first place.
Strategic Reality: The Rule of 40 was not a law. It was a convenient shorthand for an unusually predictable cost curve. AI inference costs turn that curve into a moving target, and UK boards using the old benchmark to judge performance are measuring against a yardstick that no longer fits.
Market growth is slowing at the same time. Many SaaS categories have reached high penetration, especially in the UK mid-market, where most functional stacks are already consolidated. Adding AI features to a mature category does not automatically create new seats. It often just raises the expected feature set at the same price point, which means cost goes up without revenue following.
The critical numbers
| Metric | Traditional SaaS | AI-era SaaS |
|---|---|---|
| Gross margin | 75-85% | 55-70% (inference drag) |
| Marginal cost per user | Near zero | Scales with usage |
| Rule of 40 achievability | Standard target | Increasingly rare |
| Reported productivity uplift | N/A | 10-25% EBITDA where transformation is real |
| Bain case revenue growth | — | 38% |
| Bain case cost growth | — | 349% |
What is really happening underneath the headline
The pressure is not evenly distributed. Three patterns matter for UK leadership teams.
Inference cost is becoming a line-of-business problem, not a platform problem. In most UK SaaS companies, AI features were bolted on by a central platform team and charged to a shared budget. That worked when usage was low. Once adoption moves past early users, the cost belongs to whichever product line drives the calls — and product P&Ls that looked healthy twelve months ago now carry a cost bucket nobody modelled.
Customers are repricing software value in real time. UK buyers have become noticeably sharper about what they pay for in the past year. Finance directors are asking vendors to itemise AI feature costs, push back on seat-based pricing when usage is uneven, and justify renewals against clearly measured productivity gains. The vendors that cannot answer those questions cleanly are losing the pricing power the Rule of 40 assumed.
The productivity tailwind is real but uneven. Bain’s 10-25% EBITDA uplift is not a projection — it reflects organisations that have genuinely restructured work around AI. Most UK software companies have deployed copilots to engineering and support without redesigning the processes around them, and the savings show up as slightly faster individuals inside the same headcount, not as margin expansion.
Implementation Note: If your AI programme has not changed which roles exist, which teams own which outcomes, or which tools are retired, it is almost certainly producing activity rather than economic value. The productivity gains show up at the organisational level, not the individual one.
Success factors separating real uplift from theatre
- Process redesign precedes tool rollout. Teams that mapped the end-to-end workflow before introducing AI capture the efficiency. Teams that layered AI on top of existing process captured noise.
- Inference budgets sit with product lines, not platform. Cost visibility at the P&L level changes which features ship and which get killed. Hidden costs produce hidden damage.
- Pricing is revisited, not just packaging. Outcome-based, consumption-based, or hybrid pricing tends to outperform pure seat-based models when AI features drive the value.
The strategic choice UK boards now face
Bain frames the decision as binary: financialise the business, or invest to grow. In practice, most UK software boards are sitting in a muddled middle — talking about AI investment publicly whilst cutting it quietly to protect quarterly margin. That is the worst of both options.
Financialise the business. Cap AI investment at a level that preserves margin, accept slower growth, treat the business as a cash generator, and return capital to shareholders or investors. This is a defensible choice for mature, category-leading SaaS businesses with sticky customers. It becomes a trap for mid-market vendors whose differentiation was already thin.
Invest to grow. Accept three to five years of margin compression, fund AI meaningfully across product and operations, and aim for a stronger competitive position on the other side. This is what AI-native competitors are doing by default, because they have no legacy margin to protect. The risk is that public or PE investors punish the near-term numbers before the strategy pays off.
Critical Context: The Rule of 30 — a combined growth-plus-margin target of 30% during the investment phase — is Bain’s suggested compromise. It is a permission structure for boards to signal investment discipline whilst actively spending. UK boards that want to invest should consider framing it this way with investors now, before weak quarterly results force the conversation on worse terms.
Stakeholder impact of the decision
| Stakeholder | Financialise path | Invest-to-grow path |
|---|---|---|
| Public-market investors | Predictable, short-term friendly | Need narrative reset |
| PE sponsors | Cash yield, clean exit | IRR pressure unless hold extended |
| Customers | Feature stagnation risk | Better product, possibly higher price |
| Engineering talent | Reduced ambition, retention risk | Harder work, stronger mission |
| Competitors (AI-native) | Easier to displace you | Harder fight, clearer position |
What UK software leaders should do before the next planning cycle
The decision does not need to wait for the next board strategy offsite. It needs to inform budget allocation in the current cycle, which for most UK software companies begins now.
1. Rebuild the cost model with AI variable costs made explicit. Finance teams should separate AI inference, model access, and agent infrastructure from the rest of cost of goods sold. Until you can see those costs per product, per customer tier, and per usage pattern, you cannot price, cannot forecast, and cannot negotiate with model vendors.
2. Reset the investor narrative before the numbers force it. Whether the target audience is the board, a PE sponsor, or public markets, the framing matters. “We are sacrificing margin to build durable AI advantage” lands very differently from “our costs got away from us.” Get the story aligned with the strategy and communicate it proactively.
3. Choose one to three features where AI genuinely changes the job, not ten where it assists. Broad AI rollouts inside existing products produce high inference costs and marginal customer value. Narrow, high-conviction bets where AI performs a task the customer used to pay humans to do justify their cost and can support outcome-based pricing.
4. Treat pricing as a first-order strategy question. Seat-based models are increasingly mismatched with AI value delivery. Hybrid pricing — base seat plus usage, or outcome fees on specific workflows — is being adopted by category leaders and should be tested in renewal cycles rather than held back for a pricing project next year.
5. Build AI cost accountability into product-line P&Ls. Inference budgets sitting with a central platform team create the marketing-tech case Bain cites: costs grow 349% and nobody owns the problem until it shows in the group P&L.
Priority actions by maturity level
Early-stage UK SaaS (below £10m ARR): Use AI-native architecture choices to start with lean inference costs. Do not let the growth narrative obscure unit economics. Investors are scrutinising gross margin on AI features specifically.
Scaling UK SaaS (£10m-£100m ARR): This is where the squeeze is tightest. Run the cost-model rebuild in this quarter. Pick the investment path explicitly and make sure your board and sponsor understand it. Mid-market buyers are the most aggressive about repricing AI value.
Mature UK software (over £100m ARR or listed): Decide whether to financialise or invest — do not drift. If investing, consider preparing the Rule of 30 narrative with investor relations before the first quarter of margin compression lands.
Hidden challenges most plans miss
Vendor lock-in on model access. Switching between frontier model providers sounds simple until you measure the real cost: prompts tuned for one model degrade on another, evaluation infrastructure is provider-specific, and negotiated pricing evaporates. UK software companies should build provider abstraction early, even if it feels premature.
Data residency and enterprise sales. UK enterprise and public-sector buyers increasingly require UK or EU data residency for AI features. Software companies routed through US-hosted inference lose deals they would have won on product merit. The cost of compliant infrastructure belongs in the planning assumption, not in the next phase.
Productivity theatre in the engineering team. Copilot adoption often produces more code, not better outcomes. Engineering leaders measuring commits or pull requests as productivity proxies are almost certainly reporting false tailwinds. The honest measure is cycle time to shipped customer value, and it usually tells a less flattering story.
The AI-native competitor who reprices the category. A well-funded AI-native entrant does not need to match your feature set. They need to price below your fully-loaded inference cost, reach customers who have already accepted that AI will change their workflow, and let your seat-based model look dated. Several UK SaaS categories are one quarter away from this happening.
Warning ⚠️: The biggest strategic risk is neither over-investment nor under-investment. It is inconsistency — announcing a growth strategy publicly whilst managing the business for quarterly margin internally. Investors, customers, and talent all see through it within two cycles.
The strategic takeaway
The Rule of 40 is not dead, but it has become an output rather than a target. UK software boards should stop using it as the organising metric and instead model their business in two layers: the core, mature product running on traditional SaaS economics, and the AI-augmented layer running on different unit economics that must be measured, priced, and managed separately. Treating them as one business produces the muddle Bain describes — and the muddle is how competitors win.
The companies that will do well through the next three years share three characteristics. First, they have made an explicit, board-level choice between financialising and investing, rather than drifting. Second, they have rebuilt their cost visibility so AI variable costs are understood at the product-line level, not hidden in platform overhead. Third, they have repriced deliberately, matching the value delivery of AI features with how they charge for them.
For UK leaders, the next step is smaller than the strategic framing suggests. Get the AI cost model onto a single page this quarter. Share it with the board. Use it to choose which of the two paths you are actually on. The strategy conversation becomes much easier once everyone is looking at the same numbers.
Take Action: If your board cannot see AI inference costs broken out from cost of goods sold on a single page this quarter, that is the first problem to fix — before the strategic choice, before the pricing review, before the investor conversation. Clarity on cost is the prerequisite for every other decision.
Next steps checklist
- Separate AI variable costs in management accounts this quarter
- Decide financialise-versus-invest at board level, explicitly
- Reset investor/sponsor narrative before quarterly results force it
- Move inference budgets to product-line P&Ls
- Pilot hybrid or outcome-based pricing on one renewal cohort
- Review vendor lock-in and data residency assumptions
Source citation and attribution
Primary source: Lipman, D., Callahan, G., Goetz, D. and Sunderland, G. (2026) AI Brings Headwinds and Tailwinds to the Rule of 40. Bain & Company. Available at: https://www.bain.com/insights/ai-brings-headwinds-and-tailwinds-to-the-rule-of-40/
About Resultsense: Resultsense provides UK business leaders with strategic analysis of AI developments, translating technical shifts into board-level decisions. If you would like an independent view on how AI is reshaping your software economics, get in touch.