September 20, 2026
Every other week we’ll provide updates on the latest value levers and trends operators are asking us about in Technology and Software. If there are things you want to hear more about – shoot us a note.
We have spent the past year taking apart the software company.
We wrote about how to build an AI product that stays differentiated, how to monetize AI, how to use M&A to move faster, how to redesign the SDLC, how Customer Success needs to change, and how to manage cloud and token costs.
We also wrote about how all of this ultimately shows up in valuation and the Rule of 40.
Each article addressed one piece. Now it is time to put the company back together. Because most software companies no longer have an AI experimentation problem. They have pilots, copilots, agents, steering committees, and exciting demos. What many still lack is an integrated plan for how AI changes the business.
Software CEOs need to manage AI as a company-wide refactor: What will we sell? Why will customers buy it from us? How will we monetize it? What should we acquire? And how must the company change to build, sell, serve, and operate differently?
Here is the plan.
Start with growth, then follow the value through the company
The AI Refactor has two sides.
The first is the growth engine: how AI changes what you sell, the value you create, and the markets you can serve.
The second is the operating engine: how AI changes the way you build products, go to market, serve customers, and run the company.
You need both. In most cases, you should run them at the same time. But they are not equally important.
Product strategy and monetization represent a relatively small share of OpEx. Yet those decisions govern essentially all the revenue. Product and engineering, GTM, and G&A represent much larger cost pools, but those functions execute against the direction set by the product and commercial strategy.
Translation: Do not confuse where the money is spent with where the value is decided.
Software companies are still valued primarily on the quality and durability of their growth. As we discussed in The 3 Questions Every Software CEO Needs to Answer, investors want to understand whether a company can protect retention, build a credible AI product strategy, and translate AI into measurable productivity.
Efficiency matters. But cutting G&A will not rescue a company whose product is becoming irrelevant.
Start with growth.
Refactor the growth engine
The growth side has three parts: product strategy, monetization, and M&A.
1. Build something customers cannot easily recreate
Models will change. Infrastructure will improve. Competitors and customers will have access to many of the same tools. Ask yourself: What will make our product uniquely valuable when everyone has similar AI?
As we wrote in If Everyone Has the Same AI, What Makes You Different?differentiation will increasingly come from what sits around the model:
- Proprietary data and context
- Deep understanding of customer workflows
- Feedback loops that improve the product with use
- Evaluation systems customers can trust
- Embedded expert judgment
- The ability to take action inside critical systems
The product strategy has to start with the future customer workflow, not the latest technical capability.
How will the customer’s job change over the next two to four years? Which activities disappear? Which decisions become more important? Who performs the work—a person, an agent, or some combination? Where does your product sit in that workflow?
These choices determine the talent, engineering investments, data, partners, and portfolio you need.
Product strategy is not another AI workstream. It is the strategy.
2. Monetize the value, not the technology
Once you know what value the product creates, decide how to get paid. This is where the discussion often collapses into “seats versus consumption.” That is too simplistic.
The first question is whether you are charging against the right meter. Depending on the product, the answer might be seats, usage, agent actions, transactions, resolutions, outcomes, or a hybrid of access and consumption.
In How Leaders Are Actually Pricing AI, we laid out the core decision tree. Does AI enhance the product, automate part of the customer’s workflow, or commoditize functionality previously sold through seats? The answer should shape the meter, packaging, adoption strategy, and pace of monetization.
This gets even more interesting as products expose APIs and MCP servers. Should MCP access be bundled to drive adoption? Metered by action or consumption? Packaged for developers and partners? Used as a distribution layer that increases the value of the underlying application?
There is no universal answer. But “we will figure it out once customers start using it” is not a monetization strategy. Define a clear value unit, instrument usage, avoid billing surprises, and understand the cost to serve each unit before scaling aggressively.
3. Buy speed when building will take too long
The third part of the growth engine is M&A. The traditional M&A process begins with market maps, financial screens, synergy cases, and a spreadsheet projecting an orderly future.
AI is not feeling particularly orderly.
The more strategic question is: Where is buying speed more valuable than building patiently?
An acquisition can provide proprietary data, a differentiated capability, scarce technical talent, workflow expertise, or distribution into an important customer base.
Our Operator Spotlight on how Zendesk used M&A showed how a software leader used multiple acquisitions to assemble capabilities across automation, quality, analytics, search, and voice, then integrate them into a broader platform strategy.
The lesson is not to buy every AI start-up. It is that M&A can be part of the product-development system.
As we argued in You Can’t Spreadsheet Your Way Through a Technology Transformation, management teams may need to place more weight on strategic relevance, capability ownership, and time-to-market.
Sometimes, an acquisition is essential to reaching the required destination fast enough, regardless of the asset’s standalone return.
Differentiate the product. Monetize the value. Buy speed where building will be too slow.
Refactor the operating engine
Once you are clear about where the company is going, redesign how it gets there.
1) Product and engineering: Turn faster builds into greater customer value
Product and engineering often represent roughly 30% to 40% of OpEx. It is a large pool, and the work is highly applicable to AI.
But the goal is not simply more code. Coding sits inside a larger system of discovery, architecture, specifications, testing, security, deployment, and iteration. If coding accelerates while everything else stays the same, you have only moved the bottleneck.
The next model must rethink PDLC, SDLC, and CI/CD together. Engineers may spend less time drafting code and more time defining systems, validating outputs, managing agents, and resolving ambiguity. Precise specifications, QA, security, and architecture become more important.
We went deeper on this in The Agentic SDLC. The CEO question is what the company will do with the capacity. Will it release faster? Address technical debt? Increase experimentation? Improve quality? Reduce cost?
“Engineering is more productive” is not an outcome until leadership decides where that productivity goes.
2) GTM: Turn productivity into growth
GTM often represents roughly 25% to 35% of OpEx.
AI can improve targeting, research, content, lead qualification, seller preparation, implementation, support, and Customer Success.
That is a lot of opportunity. It is also a lot of opportunity to create more activity without better results.
More personalized emails are not a growth strategy.
The prize comes from redesigning the commercial system. Better ICP intelligence should change marketing spend. More productive reps should change coverage, quotas, specialist roles, and compensation. Better post-sale intelligence should change adoption, expansion, and retention.
Leadership must decide whether a productivity gain translates into more growth, better customer outcomes, lower cost, or a deliberate combination.
Customer Success becomes especially important as products move toward consumption, APIs, MCPs, and outcomes.
In a seat model, adoption matters. In a consumption model, adoption is revenue.
Someone must help customers change workflows, prove ROI, manage spend, and turn consumption into expansion. We explored the role, automation, and talent implications in Scenes from the Customer Success Reset.
3) G&A: Take the clean cash, but keep it in perspective
G&A contains structured, repetitive workflows across finance, HR, legal, procurement, and internal IT. Some percentage opportunities may be substantial.
Management teams should pursue them. In some cases, G&A can produce the fastest underwritable savings and fund investment elsewhere. But G&A is generally a smaller absolute pool, often around 10% to 15% of OpEx.
Do not mistake the easiest business case for the most important strategy.
Automating accounts payable is useful. It is not a substitute for deciding whether your product will matter three years from now.
Refactor the cost architecture
Cloud, inference, and token costs sit largely outside the OpEx map, generally in COGS. Economically, they belong inside the AI Refactor.
A company can make every employee more productive while quietly making each customer interaction less profitable.
In Umm…Things Sure Got Expensive, we covered the broader cloud playbook: architecture, capacity, routing, caching, contracts, and FinOps.
We then went deeper in Yet Another AI Cost to Manage: Tokens.
Tokens are an input into work. They can replace or enhance labor, enable new services, or burn money while an agent cheerfully goes in circles.
Leaders need visibility into cost by product, customer, workflow, model, agent, and business outcome. Track dollars per resolution, workflow completed, deployment, customer interaction, or agent action.
The AI Refactor is incomplete if labor productivity improves while unit economics deteriorate.
Manage the refactor as one system
The pieces cannot be managed separately.
A new AI product changes the pricing model. Pricing changes the role of Customer Success. Increased usage changes cloud and token costs. Those costs affect gross margin. Gross-margin pressure affects the Rule of 40. The roadmap may reveal capabilities that need to be acquired, and an acquisition changes the architecture and operating model.
Pull one thread and the rest of the company moves.
As we discussed in Rule of 55, AI may create an awkward period where investment arrives before the benefits. Product development, acquisitions, talent, adoption, and inference costs can precede the payoff.
That is not necessarily failure. But leadership must be explicit about the trade.
What are we investing? What should improve? When should it show up? Which leading indicators tell us the plan is working? What would cause us to change course?
And across every function, ask one more question:
Where did the productivity go?
If the answer is unclear, the value probably has not reached the customer or the P&L.
The CEO’s AI Refactor checklist
At the next leadership meeting, ask:
- What will make our product more valuable and difficult to replace?
- How will the customer workflow change over the next two to four years?
- Are we charging against the right value meter?
- How should we monetize APIs, MCP access, agent actions, and outcomes?
- Which capabilities would be faster or safer to acquire than build?
- How must the full product-development lifecycle change?
- Will GTM productivity become growth, lower cost, or better customer outcomes?
- What role must Customer Success play in driving consumption and proving value?
- Which G&A savings can fund the larger growth agenda?
- Do we understand cloud and token costs at the level of customer value?
- How will the plan affect gross margin, growth, and the Rule of 40?
- What are we stopping so the refactor has real capacity and accountability?
We have been building toward this plan to turn AI from a collection of initiatives into a coherent strategy for growth, productivity, and long-term relevance.
Start with the product. Follow the value through the business. Then refactor the whole system around it.
Partner & Global Leader of Software Practice
Boston
