AGI SALT, born in AI Connect with SALT

SALT / CareersOpen roles

Your judgment is the job.

AGISALT is the born-in-AI Microsoft partner that reimagines the work and deploys the AI agents that run it. We are Customer Zero: we run the model we sell before we sell it. That includes how we hire.

Agents take the volume; your people direct the work. That is what we sell, and it is how we staff. In these four roles the judgment is the work: what to build, what to refuse, and when the confident answer on the screen is wrong.

Microsoft Solutions PartnerSee the four roles ↓

How we work

Human-led, agent-operated, outcome-driven.

Agents take the volume; your people direct the work. That sentence is how we sell to customers and it is also our employment model. Every role below states what the agents handle and what the human decides.

Three of the promises we make to customers. Read them as the terms of the job.

We run it ourselves.

AGISALT operates its own business on the exact secure-agent model it builds for you. We don't sell a transformation we haven't lived: proof, not promises.

Your people stay in the lead.

AI agents take the volume; your people direct the work. You become AI-native by elevating your team, not replacing it. In these four roles, you are the person directing.

Secure, on Microsoft.

Your agents run inside Microsoft's intelligence and trust platform: governed, observable, enterprise-grade by design.

The ARC, where every engagement runs

Activate
Agree the direction.
Reimagine
Redesign the work, with AI agents in the room.
Cowork
Run your business on agents, your people in the lead.

How we hire

We publish the exercise, because the exercise is the job.

Every role below is assessed the same way. We hand you real agent output with real defects planted in it, and we watch what you do.

  1. A short conversation about work you have owned, not a competency script.
  2. The review. You get an agent-produced artifact for your role: a deployment plan, a set of process maps and requirements, a business case, or a pull request. Each one contains something wrong that reads as correct.
  3. A conversation about what you found, what you would change, and how you would say it to the person who asked for the opposite.
  4. A decision, made by a named human who tells you the reasoning either way.

Delivery Engineering

The ARC · Cowork

Forward Deployed Engineer

  • Full-time
  • Oak Brook, IL
  • On site

Forward deployed means deployed into the customer's environment: their tenant, their data, their undocumented exceptions. Our agents can produce a deployment plan in an afternoon. Making that plan true inside one real company is the hardest engineering problem we have.

What you decide

  • Build, configure, or refuse. Which of the three, and the argument for it.
  • Whether the customer's stated requirement is their real problem, and what to do when it is not.
  • When agent output is wrong in a way the customer cannot see.
  • When a data or scope problem has to stop the work rather than be engineered around.

What we are looking for

  • Strong engineering fundamentals, and the judgment to design systems that survive contact with messy enterprise data.
  • Experience building and integrating software through APIs and cloud services, on Microsoft Azure or comparable.
  • Microsoft Dynamics 365 Finance and Operations or Microsoft Dynamics 365 Business Central experience is welcome, functional or technical.
  • Comfort in front of a customer, including telling them something they do not want to hear.
  • The instinct to question output that looks confident and reads well.

You are accountable for whether it runs in the customer's production environment after we hand it over.

Where the human sits
Agents handle

Discovery and analysis, data migration, configuration, validation, and the first pass of integration mapping. Faster and more completely than any person.

You decide

What they build, whether what came back is right, and what happens when it is not. Nothing reaches a customer that you have not judged.

With your application: tell us about a time you were the only person in the room who thought the confident answer was wrong.

Apply now

Customer Outcomes

The ARC · Reimagine

Systems Thinker

  • Full-time
  • Oak Brook, IL
  • On site

Most enterprise projects meet every requirement and still fail. A requirement is an atom. A system is the whole thing: the process, the data, the exception paths, the people, the incentives, and now the agents doing the work. We are hiring the person who holds all of it at once.

What you decide

  • What the system should be, not what the customer says it is. Recording is not the job.
  • Where it actually breaks, which is rarely where the customer says it breaks.
  • The two that move it. Twenty changes will be requested. Two of them move the system. You choose those two and defend the eighteen you refused.
  • Which exceptions are genuine requirements and which are years of habit that should not survive the redesign.
  • What breaks because we fixed this. Second-order effects are yours, and nobody else will raise them.

What we are looking for

  • Real depth in enterprise operations: finance, supply chain, manufacturing, order to cash, procure to pay, hire to retire. Deep enough to tell a constraint from a preference.
  • Enough Microsoft Dynamics 365 or comparable business-applications fluency to know what is possible before you promise it.
  • The habit of following a consequence across a department boundary rather than stopping at your own scope.
  • The willingness to tell a customer, early and plainly, that the system they are attached to is why the last project failed.
  • Judgment over tool history. The tools in this role will change while you are here.
Where the human sits
Agents handle

Requirements capture, current-state mapping, fit-gap, functional design drafts, test scripts, and first-pass data mappings. They will also hand you a complete document describing a system that cannot survive a month-end close, and not know.

You decide

The design a customer signs. And what the agents do inside the system you are redesigning, what a human reviews, and how anyone can tell it is working.

With your application: tell us about a system you were asked to document and told them to change instead, and what happened.

Apply now

Customer Outcomes

The ARC · Activate through Cowork

Outcome Manager

  • Full-time
  • Oak Brook, IL
  • On site

Our contracts commit to outcomes, not to activity. Somebody has to be the named human accountable for whether the outcome arrived. Not a status report about the outcome. The outcome.

What you decide

  • Whether an outcome was genuinely met or merely delivered. These are different, and only one of them is yours to sign.
  • When to tell a customer that the target they bought was the wrong target.
  • What reaches their executive, when, and in whose words.
  • When to tell us internally that a milestone is not met. This is the part of the job that will make you unpopular, and it is the part we are hiring for.

What we are looking for

  • Ownership of enterprise customer outcomes in consulting, delivery, or customer success on complex software.
  • The ability to build a credible business case in the customer's own numbers, and to say plainly when those numbers do not support the claim.
  • Enough technical fluency to challenge an engineering answer without pretending to be an engineer.
  • Comfort delivering unwelcome news early to senior people.
Where the human sits
Agents handle

Building and maintaining the value case, engagement coordination, evidence assembly, and production support signal once the agents are live.

You decide

Outcome acceptance. That is never agent-executed, at any level of automation. A named human signs, and it is you.

With your application: tell us about an outcome you owned that did not land, and what you did about it.

Apply now

Product Engineering

In house

Product Engineer

  • Full-time
  • Oak Brook, IL
  • On site

We are building the product that makes the other three roles possible: Business Agents, the secure AI agents that implement and drive Microsoft Dynamics 365 Finance and Operations and run enterprise business processes. It is early. What you build is what ships.

What you decide

  • What becomes product and what stays a one-off. The line between the two is where our margin lives.
  • How an agent should fail. Not whether it fails. How, and what it tells the human when it does.
  • What the field learned that belongs in the product, and what is one customer's exception dressed as a requirement.
  • When your own agent-assisted code is not good enough to merge.

What we are looking for

  • Strong fundamentals in software design and systems thinking, and evidence you have shipped something real and maintained it afterward.
  • Fluency across the Microsoft stack, Azure, Power Platform, .NET or X++, or the demonstrated ability to acquire it quickly.
  • Experience with AI agent or Copilot frameworks is welcome.
  • Learning velocity. The specific tools in this role will change, and we are hiring the person who moves with that.
Where the human sits
Agents handle

Parts of your own build. You work agent-first on your own code, the same way our product asks our customers' people to.

You decide

Every line that ships under your name. Reviewing agent output critically is a core engineering competency here.

With your application: send us something you built, and tell us what is wrong with it.

Apply now

How to apply

Send a resume and a short note answering the question at the end of the role. Applications go to careers@agisalt.com.

Equal opportunity

AGISALT Inc. is an equal opportunity employer. We consider all qualified applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, veteran status, or any other characteristic protected by law.