McKinsey: AI Coding Agents Are Changing Software Buying
AI coding agents are beginning to influence what businesses choose to purchase. McKinsey found that 32% of respondents said their organisations had rejected at least one software product or feature because they could build the functionality internally.
TL;DR
- 32% of respondents said their organisations had rejected at least one software purchase in favour of building the functionality internally.
- The research covered 1,719 respondents across 97 countries.
- The finding does not mean a third of companies have stopped buying software.
- Internal tools still bring maintenance, security, compliance and reliability costs.
What McKinsey found
In its State of AI in 2026 survey, McKinsey reported that 32% of respondents said their organisations had decided against purchasing one or more software products or features. They believed the required functionality could instead be built internally using agentic coding tools.
McKinsey refers to these AI coding agents as “agentic coding tools” and elsewhere as “software coding agents”.
The online survey received responses from 1,719 participants across 97 countries. It ran from 4 May to 8 June 2026 and included different industries, company sizes and professional roles.
The results were self-reported. McKinsey did not inspect company purchasing records or independently test the software that respondents said had been built.
Who is McKinsey?
McKinsey & Company is a global management consultancy. It advises businesses and public-sector organisations on strategy, technology, operations and organisational change.
QuantumBlack, AI by McKinsey, is the consultancy’s AI-focused unit. McKinsey has published its State of AI survey series for nearly a decade, which is one reason its findings receive attention from business leaders and technology companies.
A third of companies have not stopped buying software
The 32% figure needs to be handled carefully. An organisation only had to reject one product or individual feature to be included.
The survey does not show that these businesses abandoned commercial software generally. It also does not reveal whether each internal alternative reached production, remained operational or delivered long-term savings.
The report’s wording does not clearly state that every purchasing decision happened during 2026. Describing this as “a third of companies stopped buying software this year” therefore goes beyond the available evidence.
Why coding agents are changing the decision
AI coding agents can help teams create specialised applications, automate internal processes and connect existing systems without starting every line of code manually.
That makes building more attractive when a company only needs a focused tool rather than a large software package containing dozens of unused features. It may also give businesses greater control over how the software fits their existing workflows.
However, building the first working version is only part of the cost.
Internal software brings different expenses
Custom software must be tested, secured, updated and maintained. Someone also becomes responsible when an external service changes, sensitive data is exposed or the internal tool produces unreliable results.
McKinsey also found that 20% of respondents said the cost of running AI had limited how much their organisations could use it. Meanwhile, only 37% of respondents said AI had improved their organisation’s operating profit, almost the same proportion reported in McKinsey’s 2025 survey. This suggests that although businesses are using AI more widely, most respondents still could not point to a positive company-wide profit impact.
This does not prove that internal development is a poor decision. It shows why replacing a subscription fee with development work should not automatically be treated as a saving.
My Digitalize Life View
I can see why businesses are tempted to build exactly what they need rather than pay for a software package filled with features they’ll never use. Coding agents are making that far more practical, and this 32% figure suggests the traditional build-versus-buy decision is already shifting.
However, creating something that works is only the beginning. Security, compliance, reliability, maintenance and staff time can quickly turn an apparently cheap internal tool into a false economy.
I think this is a positive development for companies prepared to take responsibility for what they build, but it doesn’t mean commercial software has suddenly become unnecessary.
What to watch next
The next useful evidence will be whether these internally built tools remain operational and whether they genuinely reduce spending once staff time, AI usage, security and maintenance are included.
Commercial software companies may also respond with lower prices, simpler products or more flexible features. Coding agents could therefore change what businesses buy even when they do not replace purchased software completely.