Saxon AI makes the back office answer back

A football coach asks when the helmets will arrive. Somewhere inside the supplier’s computer systems, the answer exists. The salesperson standing beside the field still has to find it. That small gap between possessing information and being able to use it is where Saxon AI has built an enterprise technology business.

Consider Riddell, the protective sports equipment company. Its sellers faced a familiar office absurdity: answering an ordinary customer question could require several systems, a report, or a call to somebody back at headquarters. The company had plenty of software. Information still arrived too slowly.

The answer was already there

Riddell could have expanded its sales-support operation. Its CIO, Aravind Kashyap, saw the limitation. Adding people would leave the same information bottleneck in place. Working with Saxon and Microsoft, Riddell instead built RIA, the Riddell Intelligent Assistant: a way to ask for SAP information through Microsoft Teams.

“We could have added more sales support team members. That wouldn’t have solved the problem.”Aravind Kashyap · CIO, Riddell

The interesting decision came before the technology. The project team followed sales representatives through their working routines, examining about fifty tasks. Which consumed time? Which information mattered during a customer conversation? Saxon built the AI stack; Riddell took responsibility for extracting data and governing it. The assistant emerged from that division of labor.

Saxon’s case study describes a seven-month journey to rollout, beginning with enthusiastic users and expanding regionally. Training, surveys and early-adopter testimonials accompanied the software. Reported results include 21,000-plus hours saved annually. More revealing than the large number is the habit: 73% of reps reportedly use RIA at least once or twice a week.

A business built between systems

That is a helpful entry point to Saxon itself. Led by President and CEO Gopi Kandukuri, the company works where data engineering, application development and AI implementation meet. Its buyers are organizations with existing systems, existing procedures and enough operational complexity to make a simple question surprisingly expensive.

The company describes its approach as connecting enterprise context to action. Translated into working language: give AI access to the right records, explain the business relationships those records represent, and connect the answer to something an employee can actually do. A purchasing assistant needs supplier information and approval rules. A finance assistant needs invoice records and permission to use them.

Saxon’s expertise extends underneath the conversational interface. Its data practice covers pipelines, migration, governance and analytics. Cloud and business-application services address the infrastructure and software around those data flows. Its Microsoft relationship is especially visible, although the broader proposition includes connecting other enterprise applications.

Three verbs: search, assist, act

AIssist organizes that proposition into three capabilities. Enterprise Search retrieves information across company documents and systems, with results governed by access permissions. Assistants add role-specific help inside work routines. Agent Studio provides tools to design and manage agents that execute multi-step workflows across applications.

For a buyer, the distinction matters. Finding an invoice is one task. Explaining its status is another. Updating a record or routing an exception requires additional authority and controls. Saxon sells the engineering and integration work around those steps alongside its platform. Its offering combines reusable software with services tailored to an organization’s environment.

This places Saxon in the systems-integration and enterprise-AI services market. Buyers can also assemble an internal team or commission another consultancy. Saxon’s particular pitch is the combination of its own platform and implementation practice, especially around Microsoft environments. That is a positioning claim a prospective customer can test against a real workflow.

Fifteen thousand invoices. One recurring question.

A second case makes the pattern clearer. An unnamed chemical manufacturer processed more than 15,000 invoices monthly through Payline. Validation, approvals, status checks and vendor follow-ups depended heavily on manual work. Email threads multiplied. The first thing breaking down was the movement of information between people and systems.

Saxon describes an agent integrated with Payline and Teams that reads, validates, tracks and updates invoice records. It flags discrepancies and routes complicated cases for review. Following a pilot, the deployment expanded to 200 finance professionals and more than 500 daily interactions. Saxon reports roughly 2,000 hours of monthly savings and 70% faster processing; these are vendor-published outcomes.

The transferable idea is specific: reduce the repeated handoffs around a transaction, while keeping exceptions visible. An impressive chat response accomplishes little if the underlying invoice remains stuck in the same queue.

The customer list extends beyond sales and finance. Saxon publishes work across pharmaceuticals, manufacturing, retail and financial services, often without naming the client. Its announced combination with C3IT adds digital-platform and process-transformation capabilities. Taken together, those offerings suggest a business built around recurring enterprise implementation problems rather than a single department’s assistant.

People are part of that delivery model. Saxon’s careers material emphasizes continuous learning, while its stated values include accountability and transparency. Those are employer promises; a customer can give them a concrete test by asking who owns integration failures, who reviews exceptions and who maintains the system after launch.

The work surrounding the software

There is a practical buying lesson here. Budget for data access, connectors, testing, permissions, training and continuing operations alongside the assistant. These are parts of the documented implementation, not optional decoration. A clean interface still depends on somebody owning the records underneath it.

My reading of these projects is that Saxon fits organizations prepared to do that work. Inaccessible data, unclear approval rights or employees who gain no useful time from the tool would weaken the case. A pilot should measure the old task and the new one, then test whether people return voluntarily.

The detail worth copying is the opening move at Riddell: watch the work. Find the question that keeps sending an employee elsewhere. Make that answer available at the moment it matters. For enterprise AI, a coach waiting beside a football field is a perfectly good place to start.