How Eagle Point Software launched customer-facing AI on its Zendesk help center and answers 82% of customer questions

Eagle Point Software helps architecture, engineering, and construction (AEC) firms train their people and measure their skills, through two products: Pinnacle Series for learning and KnowledgeSmart for skills assessment. With IrisAgent, Eagle Point launched customer-facing AI search on its Zendesk help center, so customers get grounded answers right where they look for help, and its agents get AI answers inside Zendesk for every new ticket.

82%
Help center AI search questions answered
84%
New tickets with an AI-suggested answer
94%
Tickets routed to the right product knowledge
IrisAgent production data, July to November 2026. Ticket figures are for November 2026.
The challenge

Two products, one help desk, one shared login

Eagle Point customers reach one support team for two different products. The questions overlap in wording but not in answers, and a generic AI that mixes them up gives confident, wrong instructions.
01

Two products in one knowledge base

Pinnacle Series and KnowledgeSmart share single sign-on, accounts, and much of their vocabulary, so a question about logging in or assigning an assessment can mean either product.

02

Customers want answers before they file a ticket

Admins and learners at AEC firms visit the help center for how-to, login, and assignment questions that the knowledge base already covers, but searching it by keyword was hit or miss.

03

Knowledge behind a login

Much of the deepest product knowledge lives in an internal wiki behind Microsoft Entra sign-in, out of reach of a public web crawler.

04

Resolutions buried in ticket comments

Fixes for recurring issues were documented in support case comments, not in articles, so neither customers nor new agents could find them.

The solution

Customer-facing AI search on Zendesk, grounded per product

IrisAgent runs on Eagle Point's Zendesk help desk and help center, and answers only from Eagle Point's own knowledge.

AI search on the Zendesk help center

Eagle Point embedded IrisAgent AI search on its customer-facing Zendesk help center. Customers ask in plain language, get a grounded answer with links to the source articles, and can ask follow-up questions.

Product-aware answers

IrisAgent tags every ticket and article as Pinnacle Series or KnowledgeSmart using Eagle Point's existing Zendesk tags, so a KnowledgeSmart question is answered from KnowledgeSmart content and never from the wrong product.

Internal wiki synced through the Articles API

Eagle Point's team built a sync that pushes its sign-in protected wiki into IrisAgent through the Articles API, so private product knowledge powers answers without exposing the wiki publicly.

Agent assist inside Zendesk

Agents get an AI-suggested answer, a case summary, and related past tickets on each new ticket, without leaving Zendesk.

Procedures from resolved tickets

IrisAgent turned resolutions documented in support case comments into grounded procedures and knowledge articles, covering recurring issues the knowledge base did not.

The results

What changed

Measured in production, July to November 2026:

82% of customer questions answered on the help center

82% of questions asked through AI search on the Zendesk help center got a grounded answer from Eagle Point content. When the answer was not there, customers were pointed to support instead of a guess.

AI answer coverage on new tickets rose from 72% to 84%

After product segmentation went live, the share of new tickets with an AI-suggested answer climbed from 72% to 84%.

94% of tickets matched to the right product

Product tagging now covers 94% of tickets, up from 70%, so answers draw only on the right product's knowledge.

Private knowledge put to work

The internal wiki and a refreshed help center now feed the same AI that answers customers and assists agents.

Why IrisAgent

Why multi-product software teams choose IrisAgent

  • Grounded, not guessed. Every answer traces to approved knowledge. When the answer is not there, the AI declines rather than improvises.
  • Built for more than one product. Product-aware retrieval keeps answers for one product from leaking into another, using the tags your team already applies.
  • Your knowledge, wherever it lives. Help center articles, private wikis through the Articles API, and resolutions from past tickets all feed one grounded AI.

Want customer-facing AI on your Zendesk help center?

See how IrisAgent answers your customers from verified knowledge, across every product you support.