Skip to content

As of

How to automate password resets and MFA unlocks without a scripting project

Password resets and MFA unlocks can be automated without a scripting project: an AI agent connected to your identity provider resolves them in Slack or Teams, checks each request against a current record of the person, and logs every step.

The most expensive password reset your service desk will ever do looks like every other one. Someone calls or messages, says they're locked out, sounds a little rushed, and asks for a new password or a fresh MFA enrollment. Attackers have made this a habit. A CISA advisory describes intruders who posed as employees to convince help desk staff to reset the employee's password and move the employee's MFA to a device the attacker controlled. They didn't break the login page. They asked the people who run it.

The usual defence is a one-time code or a few knowledge questions: send a code to the phone on file, ask for an employee ID, a manager's name, a start date. Those checks prove something narrow. A code proves someone holds that device right now. Knowledge questions prove the caller did their research, and a manager's name and start date are often on a public profile. Neither proves the request makes sense.

That is the gap worth closing before you automate anything. A reset that runs faster on the same weak check is just a faster way to hand over an account.

What tools can automate password resets and MFA unlocks without a big scripting project?

Three kinds of tools do it today: self-service reset portals from your identity provider, scripted runbooks or RPA bots, and help desk add-ons that trigger a reset from a ticket. A fourth option, an AI agent that works through your identity provider under written policy, covers the requests the first three leave behind.

Each of the first three earns its place. Identity provider self-service portals are the cleanest answer for anyone who enrolled recovery factors in advance: they prove who they are and set a new password, and nothing reaches your team. They can't help someone who never enrolled, or whose lost phone was the factor.

Scripted runbooks and RPA fill those gaps for the cases they were written for, until the directory changes, an acquisition brings a second identity system, or the author moves on. Help desk add-ons put a reset button on the ticket, which makes a technician faster but leaves the judgment where it was: one person, under time pressure, deciding whether the caller is who they say.

All three share a blind spot. They verify the requester with whatever the reset tool can see, which is usually the identity provider alone. The facts that would expose a bad request sit in other systems.

How do you verify who is asking before a reset?

You check the request against live facts about the person, not just against a code or an answer the caller can supply. A good check asks whether this reset, for this account, right now, fits what your systems know.

Verification should consider at least these:

  • Employment status in HR: is the person active, on leave, or already terminated? A reset for someone on parental leave, or for an account whose owner left last Friday, is a different request.
  • Recent device and factor changes: did their phone or MFA device change in the last day or two? A new device followed by a reset request is the exact sequence the attack relies on.
  • Account ownership: does the account being reset belong to the same person who is asking? People hold second accounts, admin accounts and old contractor accounts, and an attacker who has one identity will ask for another.
  • Where the request comes from: is it coming from the person's usual, signed-in identity in chat, or from a new address, a new number, or someone asking on their behalf?

None of these is exotic. The hard part is that no single system holds them. HR knows employment status. The identity provider knows sign-ins and factors. Device management knows when a phone was replaced. And each of those systems may know the same person by a different ID: an email address in one, an employee number in another, a contractor account that was never merged when the contractor became an employee.

So the real requirement is one joined, current record of the person: every account they hold, tied together, with each fact showing which system it came from and when it was last observed. A fact that was true last month is not good enough for a reset today. Without that record, the checks above collapse back into whatever the reset tool can see, and you're back to trusting the code.

A short example. A request arrives in chat from a contractor's account asking for an MFA reset. The identity provider shows the account active. HR shows the contractor converted to an employee three weeks ago, with a new account that has been in daily use since. The old account should have been retired, and a reset request on it is the moment to ask why someone wants it back.

What to do about lost phones and MFA unlocks

An MFA unlock is riskier than a password reset, because the second factor is what protects the account when a password leaks. Resetting it hands the account to whoever is asking.

Treat the two separately. A routine lockout from too many failed prompts can usually be cleared once the request checks out. Replacing the factor itself deserves the full set of checks and often a second person.

A second example. An employee upgrades their phone, restores it from backup, and finds the authenticator app empty. They are still signed in to chat on their laptop, HR shows them active, and the new phone is already enrolled in device management. Everything lines up, and your policy can let that request through while stopping the same words from an unfamiliar number.

When an AI agent should hand a reset to a person

Automation should know when to stop. Some requests belong with your team even when the checks pass: a reset for an account with administrative rights, a request just after a suspicious sign-in alert, several resets for one account in a day, or a request made on someone else's behalf.

When the agent hands one of these over, the person who picks it up should see why it stopped, who asked, from where, and which facts didn't line up, without starting from scratch.

How STLabs resolves resets and unlocks with an AI agent

With STLabs, an employee asks in Slack or Teams, in their own words: "I'm locked out" or "new phone, MFA isn't working." The AI agent recognizes the request, reads the person's record from Axiom™, applies the checks your policy sets, and runs the reset through a workflow your team approved.

Axiom™ is the part other reset tools don't have. It builds a live model of your organization from your identity provider, HR system, device management and cloud tools, and it syncs in real time when someone joins, changes roles or leaves. When those tools know the same person by different IDs, Axiom™ keeps one record with every ID attached. Every fact records which system it came from and when it was observed, so the record behind a decision shows not just what the agent knew but where it came from and how fresh it was.

What the agent may do is set in Policies & Approvals: policies your team writes once, enforced on every request. Your policy decides which checks a request needs, which resets the agent can complete on its own, and which wait for a person's approval, such as an MFA factor reset for an administrator. When the agent hands a request over, it arrives with the requester's identity, device, access and recent changes already attached.

The connections are prebuilt. STLabs connects to Okta, Entra ID, Active Directory and Duo among its 140+ integrations, and there is no script to write. If your team keeps resolving the same MFA lockout by hand, Workflows spots the repeated fix and proposes it as a workflow.

Everything is on the record. STLabs logs every AI decision, workflow run and configuration change, and workflows are versioned so any change can be reviewed and rolled back. When an auditor asks who reset an account and why, the answer is in the log entry.

This is where an AI-native approach differs from a faster reset button: the agent decides with the same facts your best analyst would look up, every time, instead of the ones a caller can supply.

How to test this on your own systems

Pull last quarter's password reset and MFA tickets and read twenty of them. For each, note how the requester was verified and what that check actually proved: a device, a piece of public knowledge, or a fact from your own systems.

Next, pick five of those requesters and look them up in HR, your identity provider and device management. Count how many IDs each person has across the three, and how long it took you to tie them together.

Then try the hard cases. Ask for a reset on an account whose owner is on leave. Ask for an MFA reset from a contractor account that should have been retired. Ask for one on behalf of a colleague. A good setup completes the routine request, stops on the others, and leaves a record that shows which fact stopped it. For the wider set of requests this approach covers, see How to reduce tier 1 tickets.

Start with the reset

Resets and unlocks are a good first request to hand to an AI agent: frequent, well defined, and risky enough that the rules must be written down. STLabs works alongside the service desk you already run, and teams go from integration to results in five days. The plan is in Moving to STLabs from your current service desk.

A code proves who holds the phone; only the record proves who should.

Frequently asked questions

Can password resets be automated without scripting?

Yes. Identity provider portals handle people who enrolled recovery factors, and an AI agent connected to your identity provider can resolve the rest in chat under written policy, with no scripts to maintain.

How do you verify identity before an automated password reset?

Check the request against live facts, not just a code: employment status in HR, recent device or factor changes, whether the account belongs to the person asking, and whether the request comes from their usual identity.

Should MFA resets be automated in ITSM?

Routine MFA lockouts can be, once the request checks out. Replacing a lost factor or resetting an administrator's MFA deserves the full set of checks and often a person's approval before it runs.

How does help desk social engineering work?

An attacker contacts the help desk posing as an employee and asks for a password reset or a new MFA enrollment. One-time codes and knowledge questions often pass, because they don't test whether the request makes sense.

See it on your tools.