Agentic AI may help companies build bespoke software rather than replace the software industry wholesale, argues Infobip innovation chief Krešo Žmak. The practical shift is from asking what agents can replace to deciding what is worth building in-house.
Watch Desk analysis
What happened
Žmak’s Forbes essay describes enterprises combining conventional software with agentic features where they make sense. He cites McKinsey’s 2026 survey, in which 32% of respondents said their organisation had decided against buying at least one software product or feature because it could build that functionality using agentic coding tools.
The examples are not all independent evidence of a broad trend. Žmak says his own company moved from thousands of employee experiments towards use cases with clearer returns, and cites DoorDash results as examples of measurable gains: a 49% reduction in transfers to live agents and $3 million in year-on-year operating cost savings. Those figures are reported in his essay, not independently assessed here.
Why it matters
The build-versus-buy decision could change if AI tools make custom applications cheaper and quicker to produce. But building software is not the same as making it reliable or economical. Žmak warns that agentic systems can behave unpredictably, add costs and need guardrails. His suggested test is refreshingly unglamorous: measure productivity, revenue and cost, then expand only where the numbers justify it.
Our read
This is a useful argument for getting past the demo stage. “We built it with AI” is not a business case; a better question is whether the tailored tool works better than the product it replaces, after maintenance, oversight and running costs are counted. Žmak writes as a company executive in a Forbes Council essay, so the examples should be read as his account, not a neutral industry audit.
What to watch
- Whether companies disclose what custom AI-built tools cost to maintain, not just to create.
- Whether build decisions lead to measurable savings or simply more software to manage.
- How organisations limit agents’ scope while keeping useful flexibility.
Discussion spark: If AI makes custom software cheaper to build, should companies replace more off-the-shelf tools, or will maintenance and oversight make buying the wiser choice?
Sources and evidence
- The Next Phase Of Agentic AI: Building Your Own Business Software – Forbes (10 November 2026, 16:15 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.