IgniteTech unveils AI that repairs, tests and ships enterprise software

From a report to a deployed fix

The system is designed to work from inside a product. A user describes a problem in their own words, after which AI reproduces the issue, writes a fix, sends it through automated checks and responds when the change has gone live. IgniteTech said examples include completion notices for three reports, with each notice identifying the version in which the fix was released. Two reports began as feature requests, while the third was a defect report.

Before writing a fix, the system questions the person who reported the issue until it understands the problem. It then reproduces the reported behavior itself. A repair is released only after automated tests succeed. Once the update is live, the system repeats the customer’s original steps on the production product to confirm that it works before notifying the customer.

Feature requests use the same process. IgniteTech says every step after the user’s request is performed by AI, while sign-in, permissions, secrets, payments, destructive data changes and external communications remain under human control.

Khoros platforms will be the first rollout

Khoros will initially apply the approach to its Aurora AI and Iris® AI platforms. The goal is to let community and social media customers report problems or request features and then see suitable changes integrated into the shared product.

IgniteTech CEO Eric Vaughan, who also leads Khoros and GFI Software, said the company is building a repeatable process around the AI rather than treating it as a demonstration. “I had to set the table: Prove it in real production software and then develop the framework that enables repeatability,” Vaughan said in the company’s announcement.

Two internal IgniteTech systems already use the method: KnowledgeWell, an internal knowledge base, and Anvil, an account-management system for one of the company’s largest product programs. The company says both are used daily by dozens of employees. Vaughan wrote them himself in less than two months.

IgniteTech also described a period in early September when Vaughan spent four days serving as a juror in Texas without turning off his phone. During that time, Anvil released four versions and seventeen changes requested by five colleagues.

Guardrails define what the AI can build

IgniteTech says the main control mechanism is a set of guardrails for each product. Every product has a plan defining its purpose and direction, and each request is checked against that plan before development begins. Requests that fit become features for every customer, while those that do not are sent to a person for a decision.

The approach is intended to keep each customer on one shared product rather than requiring individually maintained versions. Human operators retain control over logins and access rights, payments, data deletion and anything sent outside the company.

“The guardrails are the product,” Vaughan said. He added that the approach is still in its early stages, and that the system does not always build a feature as the requester intended and makes mistakes. “The difference is that it gets better every week,” he said.

A shift from backlogs to continuous software work

IgniteTech frames the system as a change to how enterprise customers experience software. The company says it has spent almost four years developing the approach, alongside decades of experience building enterprise software. Its announcement argues that customers could eventually expect software to address defects and requested improvements continuously rather than leaving them in aging backlogs.

The company says it has more than 30 years of experience developing business-critical software. Since reorganizing in 2023 as an AI-focused innovation organization, IgniteTech says it has developed three new patent-pending AI-powered innovations: Eloquens® AI, MyPersonas® and Adminio™ AI.

For users, the immediate implication is a proposed route from reporting a problem to receiving a verified fix without developers handling every step manually. The limits described by IgniteTech are equally central: people remain responsible for sensitive access, financial actions, destructive changes and external communications, while requests outside a product’s purpose still require human judgment.