AI could make writing code the easy part of enterprise software, according to five predictions from Simform’s CEO and CTO. Their central point: faster code generation does not remove the work of governing, testing, securing and maintaining what gets built.
Watch Desk analysis
What happened
In a Digital Journal article published on 9 November, Simform CEO Prayaag Kasundra and CTO Hiren Dhaduk set out five ways they expect enterprise software engineering to change by 2027. Their predictions cover how companies fund software, manage AI agents, measure returns and organise engineering work.
The Digital Journal article describes the executives’ views, not a measured forecast. The practical proposals are nonetheless specific: treat software as a long-lived asset; give agents accountable owners and defined limits; measure performance against real tasks and full costs; assign agents their own identities and permissions; and spend more engineering effort on specifications and safeguards around generated code.
Why it matters
The predictions point to a less glamorous but useful test for enterprise AI: does it improve the whole job, including checking and deployment, or merely produce more output for someone else to clean up? The article also highlights how agent permissions and accountability become practical engineering questions once software can take actions, not just offer suggestions.
These are Simform executives’ forecasts, not evidence that every company is making these changes or that they will deliver a return. But they offer businesses concrete questions to ask before an AI pilot becomes another system everyone is responsible for and nobody quite owns.
Our read
The most persuasive idea here is that code volume is a poor stand-in for useful software, and AI makes it an even worse one. Organisations should count the time and cost of evaluation, review and maintenance alongside the speed of generation. If the new code arrives faster than the team can safely use it, the productivity win has taken a wrong turn.
What to watch
- Whether companies fund ongoing software reliability and maintenance as core work, rather than treating launch as the finish line.
- Whether agent access is tied to individual identities, specific tasks and clear human accountability.
- Whether AI evaluations include real work and the costs of retries, review and rework.
- Whether engineering success is measured by working outcomes rather than code produced.
Discussion spark: As AI makes code cheaper to generate, should engineering teams put more effort into defining and testing constraints, even if that means shipping fewer features?
Sources and evidence
- AI could make writing code the easy part: Five ways enterprise software will change by 2027 – Digital Journal (9 November 2026, 20:51 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.