Software: How AI Changes the Economics of the Business

Expect a sophisticated buyer to ask what is changing in the business—and what management has learned. These questions are useful before you ever consider a transaction.

• Is revenue growth slowing? Is churn changing? Are customers buying fewer seats or less software?
• Are customers pushing harder on price? Is the sales cycle getting longer?
• Is revenue per employee changing? What is happening to implementation and services revenue?
• Are customers asking what they can build themselves?
• Are competitors releasing meaningful capabilities faster? Has a larger platform started offering something similar?
• How much of the product can now be reproduced more cheaply?
• What has management already changed? What worked, what didn’t, and what investment comes next?

The point is to connect an AI story to customer behavior, operating results and the capital still required—not to arrive with a perfect forecast.

Start with your own numbers

A customer renews but buys fewer seats. A deal closes, but only after another discount. Either may have an ordinary explanation. Start with what you can observe before attributing it to AI.

• Is revenue growing, flat or declining? Is growth slowing?
• Is customer retention changing? Are customers buying less?
• Are margins and revenue per employee improving or deteriorating?
• Is the sales cycle changing? Is pricing pressure increasing?
• Is headcount growing faster than revenue?

Are you seeing the early signs of AI competition?

• Are customers asking for more self-service—or doing more themselves?
• Do they want faster answers and product changes? Are they asking why something cannot be built or delivered immediately?
• Are competitors improving faster, doing more with fewer people, or including capabilities we charge for?
• Has a new competitor appeared with a very different cost structure?
• Has a customer said it can now do something itself that it used to buy from us?

Are our competitors getting better faster than we are?

Good revenue and margins can coexist with a weakening competitive position. Renewals and existing contracts may hold up while the next buying decision is already changing. Look for a pattern; one lost deal is not proof of an AI problem.

The customer can build more

A customer needs a portal that connects two systems. Buying was once the practical choice because building and maintaining it cost too much. AI can make a limited internal application economical. That does not mean the customer should build its own ERP.

Internal tools, dashboards, integrations, portals and specialized workflows are closer to the build-versus-buy boundary. What part of our product could a customer reasonably build for itself now that it couldn’t build economically before?

The vendor can build faster

Your team has the same opportunity. But faster development can become a customer expectation: faster features, customization, integrations and fixes. The customer’s question becomes: “If you can build it faster, why can’t you give it to me?”

The cost of making some software is falling

When a useful feature takes less effort to produce, more competitors can offer it. Savings can fund a better product, support lower prices or improve margins. Competition will determine how much the vendor keeps. A cheaper first version does not remove the cost of reliability, maintenance or customer acquisition.

Some software becomes easier to reproduce

Narrow features, basic reporting, simple applications and routine workflows face more substitutes. A platform may include what a specialist sells separately. The exposed product is one the customer can replace without losing much operational knowledge or taking much risk.

Some software becomes more valuable

A system of record that holds trusted data, handles critical transactions or coordinates an essential workflow can become the foundation for new capabilities. Difficult integrations, security, reliability, domain knowledge and someone accountable when things fail still matter.

These are sources of value, not permanent protection. The boundary can move. Ask what customers would actually lose by leaving—and whether that answer is getting stronger or weaker.

The examples below show different economic pressures. Announcements reveal decisions and intentions; they do not, by themselves, establish savings or investment returns. Evidence reviewed through October 1, 2026.

Customers are choosing to build some things

McKinsey’s 2026 survey reports that 32% of respondents’ organizations had decided against buying at least one software product or feature because they could build it internally with AI coding agents. This is reported buying behavior, not audited proof that internal development was cheaper over its lifetime. For an owner, the takeaway is specific: find out which purchases customers now reconsider. McKinsey, State of AI 2026.

Atlassian is reallocating people and money

In March 2026, Atlassian announced a reduction of about 10% of its workforce, roughly 1,600 people, to help fund AI and enterprise sales and strengthen its financial position. Management described a changing mix of skills, not a simple one-for-one replacement of people by AI. The economic lesson: pursuing growth may require difficult reallocations before benefits arrive. This announcement does not measure the resulting productivity gains. Atlassian, March 2026 team update.

Salesforce bought capabilities and a customer base

Salesforce completed its acquisition of Fin, formerly Intercom, in September 2026. It brought in customer-service agent technology, a specialist team and an established customer base. Even a large software company can decide that buying capabilities is faster than building everything itself. For a smaller owner, joining a platform can accelerate an opportunity; it need not signal failure. The acquisition announcement is evidence of the transaction, not proof of future returns. Salesforce, completion of the Fin acquisition.

Coding gains are real possibilities, not a uniform rate

Cursor’s June 2026 Coinbase case study reports that some teams shortened idea-to-production time from 20 days to under two. That is a vendor-published customer example involving wider engineering changes, not an independently controlled estimate of savings for every team. It shows why competitors’ release speed deserves attention. Cursor’s Coinbase case study, June 2026.

Independent evidence is less tidy. METR’s small randomized study of experienced open-source developers found early-2025 tools made the studied tasks take 19% longer. Its February 2026 follow-up says selection problems prevented a reliable estimate with newer tools. Neither result settles what today’s tools will do in your business. Test claimed gains against delivered work, quality and total cost. METR’s 2025 randomized study; METR’s February 2026 follow-up.

An internal tool still needs a production owner

RSM describes business users creating dashboards, portals and workflows quickly, then encountering security, reliability and maintenance requirements before production. This is advisory experience, not a controlled study. It helps explain both sides of build-versus-buy: customers can reproduce more, while vendors can still earn their place by taking responsibility for dependable operation. RSM, moving AI-generated applications into production.

The opportunity can be work nobody has automated

A customer already has billing, CRM and ERP software, yet someone still chases approvals and reconciles exceptions between them. Bain estimates roughly $100 billion of potential US software opportunity in this work. That is a market estimate, not revenue already won. The owner’s opportunity may be to finish a valuable job for the customer and compete for labor spending—not just sell another license. Bain, the opportunity between software systems.

Imagine a small vendor that can serve twice as many customers without doubling its support team. That would change its ability to compete. It is a possibility to evaluate, not a savings assumption to put straight into a valuation.

There is measured evidence for improved support productivity: a study summarized by NBER found roughly 14% more issues resolved per hour with AI assistance at one large software company, with larger gains among less experienced staff. It studied assisted human agents using earlier technology, not autonomous support or a guaranteed headcount reduction. NBER’s summary of the customer-support study.

• Engineering and product development: less effort on some routine work can fund more releases. Review, maintenance and new product ambitions can absorb the savings.
• Sales and marketing: research, demonstrations, proposals and content may take less effort. Winning trust, reaching buyers and standing out can remain expensive.
• Customer support and implementation: routine questions, setup and integrations may become cheaper. Services revenue can also fall if customers need less help; exceptions and complex migrations still need people.
• Software and tools: internal builds or bundled capabilities may reduce some subscriptions. New AI tools add costs of their own.
• Data, AI infrastructure and security: model usage, dependable data, monitoring and access controls can become larger ongoing costs as customers use more of the product.

Smaller companies may gain ground

If a smaller firm can develop, sell and support more with the same team, it may narrow an advantage that once required substantial scale. Customer knowledge and a focused product can count for more. But large platforms may still have stronger distribution, trusted data access, infrastructure purchasing power and more money to absorb failed experiments. Scale will not matter equally in every market.

Pricing may move with the work

SVB’s 2026 survey of more than 120 venture-backed enterprise software companies found 37% used subscription-only pricing; 26% expected to do so in 12–24 months. These are a particular group’s plans, not a forecast for every software company. Seat, usage and outcome pricing produce different economics. Track revenue per employee alongside retention and gross margin after AI delivery costs. SVB, State of Enterprise Software 2026.

A company may have a clear opportunity and still lack the resources to pursue it. The owner-level question is: Can this company fund and execute the changes the market will require?

• Capital: enough for development and continuing operations before returns arrive, with room for another round of investment.
• People and product development: the judgment to choose valuable problems, ship dependable products and keep serving current customers.
• Data and systems: usable information, appropriate access and a product foundation that can support what customers expect.
• Management capacity: time to make decisions, assess results and stop work that is not paying off.
• Speed and room to experiment: the ability to learn, absorb failures and keep investing as technology and competitors change.

A convincing answer connects resources to customer demand and a plausible commercial return. “We are using AI” does not answer how much investment the company needs or whether it can sustain it.

The expensive outcome is not always doing nothing. It can be getting halfway through a change, disrupting the existing business and then discovering there is not enough money or management capacity to finish.

• Under-investing while competitors move faster—or committing too much before customers show they will pay.
• Hiring the wrong people or relying on capabilities the company cannot evaluate.
• Disrupting product reliability, customer service or the sales effort that funds the business.
• Choosing technology or building a feature that becomes obsolete or gets bundled into a larger platform.
• Exhausting management and falling further behind while trying to catch up.

If we choose to do this ourselves, can we afford to be wrong? The answer includes cash, customer trust and the time available for a second attempt. It should inform how much risk the owner retains and whether another party should share it.

Two companies can report similar revenue and margins while representing very different investments. One can use AI to sell more and serve customers better. The other must spend substantially just to preserve its current position.

An attractive AI opportunity and a substantial AI catch-up requirement are different assets. A buyer will try to separate optional growth investment from spending needed to stay competitive.

• How exposed is the product to internal builds, cheaper competitors and platform bundling?
• How durable is the customer relationship? What workflow, data or responsibility makes switching difficult?
• How quickly are competitors improving, and what investment is needed next?
• What has management already changed? What worked, what failed, and what did the company learn?

Evidence matters more than an AI label: customer decisions, releases, retention, delivery costs and the investment still ahead. A credible account of a failed experiment can be more useful than an unsupported claim that the whole business is protected.

Independent response may make sense when the company has the capital, people, management capacity, a clear commercial path and the appetite to execute. The owner can see what needs to happen and sustain investment while the business learns.

If the opportunity requires substantially more funding, talent, capabilities or distribution than the company can obtain on reasonable terms, the decision changes. Independence remains a choice with costs and risks—not an automatic measure of success.

Compare what the company can realistically achieve alone with what it could achieve with someone else. Include the owner’s own willingness to finance and lead the next phase.

• Bring in outside expertise when the gap is specific and management can direct and evaluate the work.
• Bring in capital when the plan and team are credible but funding limits execution.
• Acquire capabilities when a team, product or customer base can fill a gap and the company can integrate it.
• Partner when another business brings technology, access or distribution without a change of ownership.
• Join a larger platform when shared resources and a broader offering improve the opportunity.
• Sell when a new owner is better positioned to fund and lead the next phase.

Selling is not failure. An owner may have built an excellent company and decide that its customers, employees and future would benefit from resources another owner can provide. The right alternative depends on the actual gap, the terms available and how much ownership and responsibility the owner wants to retain.

Price matters. So does whether the buyer can deliver the future on which part of that price—or your continuing investment—depends.

Buyer capabilities vary. Salesforce’s Fin acquisition illustrates a strategic buyer adding a specialist product and team. In April 2026, Vista and Google Cloud announced shared engineering and cloud support for portfolio companies, naming a Duck Creek insurance agent as an example. That is evidence of an organized capability, not independently measured returns or a promise that every acquisition will receive the same resources. Vista and Google Cloud, April 2026 partnership.

• Can this buyer make decisions quickly and fund investment after closing?
• Which relevant people will actually work with our company? What capabilities do they bring?
• Where have they implemented AI already? What changed in customer outcomes, revenue or cost?
• How much autonomy will management have? What happens when an initiative fails?
• What will they commit to our product, data and team—not just describe at the platform level?

Who is actually better positioned to take this company through the next phase? Ask for specific examples and speak with management teams that have worked with the buyer.

This matters especially with an earnout, retained or rolled equity, or continued management involvement. Your outcome can depend on post-closing budgets, priorities and execution. Understand who controls those decisions and how the agreed terms address them.