MacPaw pitches IT teams an AI help desk built on a bot that resolved 30% of its own tickets

Bradley Chambers has been an Apple IT admin since 2009 and writes 9to5Mac's Apple @ Work column ^[14]. He describes a failure IT teams already know. An employee asks the internal AI tool a question, and it answers from whatever it can find, including old product docs, outdated policies and random Reddit threads ^[11]. MacPaw's 2026 AI at Work report found that 90% of companies, all but about one in ten ^[2], deploy AI without thoroughly auditing their internal knowledge base first ^[2].

Both trust figures in Chambers' column come from MacPaw's own survey ^[2][3]. The column does not include the survey's sample size, Leebry's price, or a customer outside MacPaw.

At MacPaw, employees asked the same questions repeatedly, even when the answer was already written down in docs, intranets or Slack threads ^[4]. Leebry's response is a source link on every answer and a way for employees to flag stale material ^[7].

MacPaw calls Leebry a Work AI platform ^[1]. Under the label are two jobs an admin already has on the list: level 1 tickets and access changes ^[6]. "Businesses don't need another tool to check, they need work taken off their plate, without sacrificing oversight or control," said Dan Jaenicke, Director of B2B Product Development at MacPaw ^[12].

The only deployment result on record is MacPaw's own, measured on MacPaw's documents and staff. Within a few months of the 2025 hackathon, the bot was resolving 30% of the company's level 1 IT tickets ^[5], leaving 70% unresolved by the bot ^[1].

Chambers raises the trust question himself: why an IT team should trust a company that spent nearly two decades on consumer Mac software such as CleanMyMac to build enterprise tooling ^[1]. "I think the answer is actually in the product's origin story," he wrote ^[13]. He is more specific about access. At SMBs without a dedicated IAM team, he wrote, access changes are mostly manual and slow, and departing employees keep access that should have been revoked days earlier ^[10].

Two axes sort the decision: whether a named person edits the docs when they are flagged, and whether you have a dedicated identity team. A team with a doc owner and no identity team is the SMB Chambers describes, and I'd run a pilot there. Where both exist, provisioning overlaps work someone already does, so test only the ticket side. Teams with neither should look at the access automation first and expect answers only as current as their docs. An identity team without a doc owner should wait. The tradeoff in the strongest quadrant is becoming an early outside customer of a company whose published result is its own help desk ^[5]. Before any pilot, count last month's level 1 tickets whose answer already existed in writing. That count is roughly the ceiling for a bot that works by pointing to sources, and it is the figure to hold against MacPaw's 30% ^[5].