Decimal's Eightfold veterans raise $4M to keep support tickets away from engineers

Why it matters

Decimal is betting that AI will move technical support beyond answer retrieval and into debugging. If it works, software companies can scale support without routing every hard ticket to engineering.

Sanjeet Hajarnis and Kevin Raji Cherian's Decimal AI said on September 15th that it raised a $4 million seed round for software that gives support teams access to the evidence engineers use to debug customer problems. Khosla Ventures and Kearny Jackson co-led the financing, with Atlassian Ventures and Weekend Fund participating.

For co-founders Sanjeet Hajarnis and Kevin Raji Cherian, the product grew out of a problem they had watched from inside large machine-learning organizations: a customer reports that something broke, support receives an incomplete description, and engineers must reconstruct what actually happened from code, logs, configuration and production data.

The pair met while working at Eightfold AI. Hajarnis previously worked on Facebook's News Feed ranking systems, Uber pricing infrastructure and AI products at Eightfold, according to Decimal's announcement. Cherian worked at Databricks and Eightfold, where Decimal says he built vector-search technology and infrastructure used to match candidate profiles with jobs. A Databricks patent issued in April lists Cherian among the inventors of an automated pipeline for generating embedding vectors.

Decimal began with the usual amount of ceremony for an enterprise AI startup: two people working from a living room. Hajarnis wrote that he and Cherian spent their first months exploring ideas from Cherian's home before moving the four-person team into a San Mateo office. Decimal says its product launched in March 2025.

Giving support the evidence

Decimal's platform connects to repositories, logs, documentation, configuration, production data and customer histories. When a ticket arrives, its AI Support Engineer can investigate the issue, draft an answer, take approved account or billing actions and prepare a bug fix for engineering review.

That technical scope is the core of Hajarnis and Cherian's pitch. Conventional support automation commonly retrieves an answer from documentation or a knowledge base. Decimal is built for cases where the documented answer is incomplete because the product behaved differently for one customer, environment or configuration.

"The answer to customer questions is usually sitting in what the product actually did," Hajarnis said in the funding announcement.

Decimal works inside systems including Slack, GitHub, Zendesk, Linear, Pylon, Jira, Intercom and Salesforce. Decimal says connections are read-only by default, customer data is not used to train foundation models, and customers can use private storage or virtual private cloud deployments. Those controls matter because the product's usefulness depends on access to some of a software vendor's most sensitive operational systems.

Decimal AI says the financing has already gone toward expanding the range of issues Decimal can investigate and adding integrations, according to the announcement. Its hiring plan shows an equally pressing task: convincing support, security and engineering leaders to approve a product that requires deep access and considerable trust.

Atlassian Ventures' participation fits that distribution problem. Jira sits directly in the support-to-engineering handoff Decimal wants to compress, while the product also connects with Atlassian's broader developer workflow.

Customer results come with company math

Decimal named Granola, Resilinc, Tealium, BuildOps and Lucidworks as customers, along with an unidentified Fortune 5 technology company. It says the number of support interactions resolved through the platform grew 15-fold from the beginning of 2026 through the September announcement. Decimal did not attach a starting ticket count to that figure, making the size of the underlying workload impossible to judge from the multiplier alone.

Its customer case studies provide more operational detail. Granola's support team handles twice the ticket volume and resolves 70% of common questions in chat before they become tickets, according to a case study published by Decimal. The system reads CloudWatch logs, traces relevant code and places an investigation report inside Granola's existing Plain workflow. It can also generate pull requests when a resolved issue exposes missing documentation.

Decimal also credits its deployment at Resilinc with reducing mean time to resolution by 62%, from 6.5 days to 2.5 days. Decimal's fuller Resilinc account says that improvement came during a broader overhaul that included a move to Freshdesk, better ticket management and changes to knowledge management. Decimal was one component of the result, even if the funding announcement gives it the headline.

The case studies remain vendor-published evidence. They show how customers are using the product and where the claimed savings come from, while leaving Decimal's commercial progress harder to measure. Decimal AI has supplied no revenue, pricing, retention or customer-spend figures in the announcement.

A narrower wedge in a heavily funded market

Customer service agents already command some of the largest checks in enterprise AI. Sierra raised $350 million at a $10 billion valuation in September 2025 for a broad customer-experience agent platform. In February, Resolve AI announced a $125 million Series A for agents that investigate and operate production software.

Decimal is entering through a narrower door: deeply technical customer issues that fall between a help desk and an engineering incident. That boundary is expensive because it pulls high-cost engineers away from product work, yet it also demands more context and judgment than a standard support bot can provide.

Hajarnis and Cherian are calling the layer "customer engineering," betting that support teams will gain the tools to investigate and resolve technical failures directly. The category name may or may not survive. The workflow already exists inside software companies, usually spread across tickets, Slack threads, dashboards and hurried requests to engineering.

The $4 million round gives Decimal room to turn that improvised process into a product. Its founders now have to prove that access to code and production evidence produces reliable resolutions across customers, rather than impressive demonstrations inside a handful of carefully integrated accounts.