London – November 04, 2026 -- IgniteTech is rolling out AI that takes a customer's bug report, fixes the cause, tests the fix, deploys the update and notifies the customer, with no developer in the loop for routine changes. CEO Eric Vaughan presented the approach at the Financial Times Future of AI Summit in London.
Khoros platforms get the AI-run development model first
Khoros, which Vaughan also leads alongside GFI Software, will apply the approach first to its Aurora AI and Iris AI platforms. Community and social media customers will be able to report a problem or request a feature and see it built into the shared product. Vaughan's target: a bug reported today should be fixed by the next morning.
Every fix is reproduced, tested and re-verified on the live product
Before writing a fix, the system interviews the reporter until it understands the issue, then reproduces the problem itself. A fix goes live only after passing automated tests. The system then repeats the customer's original steps on the live product before declaring the issue resolved. Feature requests follow the same path.
Two internal systems have run on the model, including four releases during a jury duty absence
KnowledgeWell, an internal knowledge base, and Anvil, an account management system for one of IgniteTech's largest product programs, already fix and improve themselves. Dozens of employees use both daily. Vaughan wrote both himself in under two months.
In early September, Vaughan spent four days on a Texas jury with his phone off. Anvil shipped four releases and seventeen changes requested by five colleagues during that time.
Blueprint checks and human controls keep each product to a single version
IgniteTech addresses the risk of self-building software drifting off course with guardrails. Every product carries a blueprint of its purpose and direction, and each request is checked against it before anything is built. Requests that fit become features for all customers. Those that do not go to a person to decide. Each product stays one product for every customer, with no custom versions to maintain.
People retain control over logins and access, payments, data deletion and anything sent outside the company. "The guardrails are the product," Vaughan said. He said the company spent nearly four years building the harness on top of decades of enterprise software experience.
Vaughan cites Anthropic and OpenAI as proof AI can write most of its own code
Vaughan said Anthropic and OpenAI have shown AI can write most of its own code, and that IgniteTech is bringing that to enterprise software. He expects customers to treat this behavior as standard within five years, possibly sooner.
IgniteTech acknowledges the approach is in an early phase. The system does not always build a feature the way the requester intended, and it makes mistakes. Vaughan said the system improves every week.