KLAS: At Keystone Systems, a Person Still Answers the Phone

Call Keystone Systems with a problem and something increasingly rare happens. A person picks up. No menu, no press one for support, no chatbot trained to pass as human. The voice on the line has a name, and the customer usually knows it already.

"We can show you a picture and say, this is the person you talk to," says Andrea Callicutt, who has spent more than twenty years at the company.

That commitment reads as sentimental only until you understand what Keystone builds, who depends on it, and where it is headed next.

Champagne taste, beer budget

"When the company was started, it was called data processing, not IT," recalls Mitake Burts, whose parents founded Keystone in the early 1980s. They had come from large organizations with entire departments devoted to computing, and saw that small companies needed the same capability without the budget for it. Delivering it meant learning everything, from speccing machines to shielding telephone lines so a lightning strike would not fry a client's computer. "These are all things that are really just kind of known now," Burts says. Back then, they were hard-won.

The library work arrived through a friend at the swimming pool who worked for IBM. He had a project too small for his employer, a client with champagne taste, and, as Burts puts it, a beer budget of around $250,000. Too small for IBM sounded very good to a startup. The client was a library serving patrons requiring accessible reading materials such as large print, Braille or audio/talking books. It wanted an integrated library system that could track reading histories and preferences and match books to patrons, a brief that current CEO James Burts describes as a library blended with a warehouse blended with Netflix. Keystone was the only bidder willing to deliver at that budget.

When big software vendors later began pulling support in-house and the reseller road narrowed, the family looked for something it owned outright and could stand behind indefinitely. KLAS, the Keystone Library Automation System, was the answer. "There's power in being a subject matter expert in a particular area," Mitake Burts says. "You don't have to be all things to everybody." Keystone now serves roughly three-quarters of the country's libraries for the blind and print-disabled, operating within the National Library Service, the federally mandated program housed in the Library of Congress. Keystone's first library customer, signed in 1983, is still a customer today. Alongside that sits a second market, statewide instructional resource centers that supply accessible educational materials, where KLAS tracks each expensive item from order to student to return, and helps centers prove state and federal government resources and grant monies are spent responsibly such as expensive materials being re-used rather than newly purchased. That is where the stakes turn personal: an elderly reader receives an audiobook in the mail, a student gets the adaptive material that makes a math class possible.

The expertise compounds. The North Carolina library ran its collection from a single-story warehouse with ample staff. Massachusetts operated out of a 19th-century building across five floors with no freight elevator, only a dumbwaiter, and shelving reached through classrooms. Solving both taught Keystone that while many libraries share common needs, none share the same circumstances, and lessons learned in one state become answers offered in another. In many cases, the company now holds more institutional memory than customers whose entire staff has turned over.

Accessibility as a continuum, not a checkbox

What most technology companies miss, Mitake Burts argues, is that accessibility "really exists on a continuum." A low-vision user who remains highly visual works nothing like someone who navigates linearly by audio, and when a person lost their sight shapes how they interact with everything. So Keystone designs every screen to be navigated by screen reader, braille output, magnification, or keyboard alone, then goes further: its specialists sit with individual employees, learn their exact workflow and assistive technology setup, and configure the system around them. The question, she says, is never just whether someone can reach the information. It is whether they can reach it efficiently. As for third-party overlays promising instant compliance, Callicutt has tested them herself. They do not work.

The human case, made practical

Keystone's caution on AI is not reflexive. The company is weighing tools that speed development, though Mitake Burts notes the risk of code that starts fast and becomes a black box nobody can maintain. Her keener concern is data. Keystone's customers are government agencies fielding constant privacy inquiries, and once information enters an outside tool, she says, one product can quietly use another beneath it, leapfrogging in ways nobody predicted.

But Keystone's concerns are not confined to how AI is developed or how data is handled; they also extend to the role it should play in direct interactions with the people its customers serve.

There is then the question of who is on the other end of the line. The median patron of a talking book library is 75. "Things I might forgive a chatbot for getting wrong when I'm changing a hotel reservation are not the interactions our customers thrive off of," Burts says. "Their value add to their community is being a human point of contact." Callicutt, who holds a master's in library science, adds the librarian's objection: machine output draws on derivative sources and provides little or no credit, in a field built on attribution.

That expertise runs throughout Keystone, with 42% of its staff holding master's degrees in library science extended across multiple departments. This gives the company an unusual depth of understanding, ensuring that decisions made across the business are informed by the realities of libraries and the patrons they serve.

The next shelf

Keystone has largely saturated its niche, and its next market is hiding in plain sight. Associations and membership organizations sit on decades of newsletters, conference recordings, proceedings, and archives with no good way to organize any of it. The fashionable answer is to hand it all to AI for automated summaries and full-text search. Keystone's answer is this fails precisely where it matters: a system with no grasp of research principles will mix historical documents with current best practice and blur industry-specific terms into their general-use lookalikes. Associations understand their resources and their members better than anyone. Keystone's library expertise connects the two, curated collections and member profiles that serve back exactly what each person asked for, positioning the association itself as a more reliable authority than a general-purpose chatbot.

"How do we make it easier for people to pull their information together," Mitake Burts asks, "while helping our customers be the information experts for their communities?" Forty years in, the answer has not changed. A person will answer the phone. The judgment stays human. And that now looks less like a limitation than a strategy.