As of
How to reduce Tier 1 service requests with AI: remove the lookup, keep the help
Reducing Tier 1 service requests with AI, in IT, HR, Facilities, Finance or Legal, starts with the lookup that comes before every fix, which an AI agent can only skip if it already knows who is asking.
An AI agent in Slack that resets passwords and hands out licenses is now an ordinary purchase. The pitch rarely changes: put the agent where employees already ask, and Tier 1 volume moves to it. Ticket counts shift a little; the analysts' day doesn't.
The usual Tier 1 model deserves some credit. A first line that takes every request, works a known fix and passes on what it can't solve is a sensible design. It gives employees one front door and IT control over who changes what. IT is where most teams see it, but HR, Facilities, Finance and Legal run the same kind of first line.
But the work in Tier 1 was never hard. It's slow because every request starts with a lookup, and a chat window doesn't make the lookup go away. Take the lookup out and the requests that land in Tier 1 today can be resolved where the employee asked, in Slack or Teams.
How do you reduce Tier 1 service requests with AI?
Take away the lookup that starts every request. Today the analyst starts with a message and a display name, and assembles the rest before a known fix can be applied safely.
Take one ordinary access request apart. Someone posts in Slack: "Can I get Salesforce? I moved to sales from customer success last month." Here is what an analyst does before anything changes.
- Who is this? The Slack display name has to be matched to a person in HR. A shared first name, a nickname or a changed surname each costs a search.
- Which of their accounts? Plenty of people have more than one: a primary login, an older account from when they were a contractor, an admin account, an alias in Microsoft 365 that doesn't match the identity provider. The license has to land on the right one.
- What role and manager, per HR? Not per the request form or last quarter's org chart. The move to sales is the fact that matters, and HR is where it's current.
- What do they already hold? A check in the identity provider and in Salesforce itself. They may still hold a customer success license that should be swapped rather than added to.
- Which policy applies? A sales license for the sales team may be standard. Someone a month into a transfer may fall under a different rule.
- Who approves? The current manager, the license owner, or nobody, because the role is pre-approved.
- How was this solved last time? Past tickets and the knowledge base show how the last few moves into sales went: which license, whether the old seat was reclaimed, who signed off. Someone has to go and find it.
- The fix. Assign the license, reclaim the old seat, add the sales groups.
Only the last step changes anything, and it should be the shortest. Often it isn't, because the fix spans systems: Salesforce, the identity provider and Microsoft 365 each have their own admin console. Where the analyst has no admin rights in one, they open a sub-ticket to the team that owns it and wait. Each lookup is minutes in a different tool, and the request waits between them. Those times are illustrative, not measured; your own tickets will give you the real ones.
None of the lookups is optional. Put the license on the wrong account and the employee still can't sign in. Send the approval to the old manager and the request waits on someone with no say over it, or gets approved by them and flagged in next quarter's access review.
Other teams' first lines have the same shape. An HR request to move someone's leave dates needs their location, the leave policy for it, their manager and how the last similar case was handled.
Why a chatbot in front of the queue doesn't remove the lookups
Walk the same request through a chat window placed in front of the same queue. The chatbot can greet the employee and ask what they need. It can't answer "which of their accounts" or "who approves" any better than the old form could, so it asks the employee, who may not know, and opens a ticket. The analyst picks it up and runs the same lookups, because answers typed into a chat aren't a source of record.
The request has a friendlier front door and the same middle, so the ticket count barely moves.
Why an AI agent's live lookup still fails
The next step is an AI agent that runs the lookups itself, live, calling HR, the identity provider and Salesforce as the message arrives. That's better, and it fails in two places.
The first is the second account. A live query asks the identity provider for the account matching the Slack user's email. It never learns that the same person has an older contractor account that still holds a Salesforce seat. The agent grants a second license, or grants on the wrong login. Joining those accounts has to be worked out across systems, ahead of time.
The second is the stale role. Systems disagree about the same person. HR shows the move to sales, the customer success group in the identity provider hasn't been cleaned up, and a manager field somewhere still names the old manager. A live lookup reads whichever system it was pointed at and acts on that. It can't tell that one answer is weeks newer than the other, so it routes the approval wrong with full confidence.
Joining identities and settling conflicting facts is careful work, and it can't be done in the seconds before a reply is due.
How STLabs gives the AI agent the lookup before the request arrives
STLabs is built on Axiom™, a live model of your organization's people, devices, apps and access, connected to the systems you already run and updated as they change. Axiom™ does the lookups ahead of time and keeps them done.
When your HR system, identity provider and device tools each know the same person by a different ID, Axiom™ keeps one record with every ID attached. So when the Salesforce request arrives, the agent already knows it's one person with two accounts, which one the license belongs on, and that the other still holds a seat. When someone changes team, Axiom™ picks up the move from HR, so the policy for the new team applies and the approval goes to the current manager. Every fact in Axiom™ records which system it came from and when it was observed, so a group membership nobody cleaned up doesn't outweigh last month's HR change.
The agent also draws on your knowledge base, past tickets and past resolutions, so it knows how your team has solved this request before.
That leaves the agent with the one step that changes something, checked against policy. Your team writes those rules once in Policies & Approvals: which requests run on their own, which need a manager or license owner, when temporary access expires. The fix itself is a workflow, versioned and reviewed before it's published, and the agent runs it through the systems you've connected. For a request policy already allows, there's no tour of admin consoles and no sub-ticket chain.
Then the decision leaves a record your team can read. Every AI decision lands in a searchable log, and the record shows which facts the agent used, where each came from and when each was observed. If an approval went to the wrong person, you can see whether HR had it wrong or the agent did.
Some requests still need a person: a policy exception, sensitive access, a laptop that won't boot. Those arrive as tickets with the requester's identity, device, access and recent changes already attached, so the analyst starts at the judgment call. That is what AI-native service management changes for Tier 1, and the same agent, workflows and Axiom™ extend beyond IT to HR, Facilities, Finance and Legal. The L1, L2 and L3 support model guide covers what happens to the tiers, and ITIL without the paperwork covers the practices around them.
Where Tier 1 volume comes from, and how to size yours
In IT, most Tier 1 volume arrives from five places: sign-in problems such as lockouts and MFA resets, access requests, how-to questions, device and app issues with a standard fix, and outage reports, where one failing service becomes hundreds of tickets. In every one the fix is already known. What varies is the lookup.
To size yours, export the last 90 days of tickets and group them by those five sources or your own categories. Sort each group into three piles: answerable from documentation, fixable by an action your policy already allows, and needing a person's judgment. Multiply the first two piles by your current handling time. That's the analyst time going to lookups and known fixes.
Keeping the employee experience intact with an AI agent
Fewer tickets is a poor result if employees feel stuck. Measure how long people wait for a fix, and how often they come back about the same request. Every handoff to a person should arrive with the context gathered, so nobody repeats who they are, and a person should always be one step away.
How to test this on your own systems
Pick twenty recent access requests. For each one, list the lookups the analyst made before the fix and the system each came from, then compare the time spent on the fix with the time spent on everything before it. If the second is usually larger, the lookup is your bottleneck.
Then count how many requesters had more than one account, and how many had a manager or team somewhere that didn't match HR. Those are the requests a live lookup gets wrong.
Finally, put three questions to any vendor. What does your AI agent know about an employee before their first message? How does it tell that two accounts belong to the same person? And for a decision it made last week, can you show which facts it used and when each was observed?
Start with the lookup
You don't need to replace the service desk you already run. STLabs works alongside it, takes on one request type at a time, and teams go from integration to results in five days. The plan is in Moving to STLabs from your current service desk.
Tier 1 was never short of people who knew the fix; it was short of the facts to apply it.
Frequently asked questions
What is the difference between Tier 1 and Tier 2 IT support?
Tier 1 handles requests with a known fix, such as password resets, standard access and how-to questions. Tier 2 handles issues that need investigation or specialist knowledge, usually escalated from Tier 1.
Does reducing Tier 1 tickets work outside IT?
Yes. HR, Facilities, Finance and Legal requests share the Tier 1 shape: a known answer that waits on a lookup of who is asking and what applies to them. STLabs extends the same AI agent beyond IT.
Why don't ITSM chatbots reduce Tier 1 tickets?
A chatbot without context about the employee has to ask who they are, which account is affected and who approves, then open a ticket. The analyst still does every lookup, just later.
How do you stop an AI agent from granting access it shouldn't?
Let it act only within policies your team writes, check each request against the employee's current role and every account they hold, and route exceptions to a named approver. Log every decision so it can be reviewed.