Product

AI Operations Assessment: What Happens in the First 30 Days

An AI operations assessment maps your processes, hours, cost and systems, then puts one workflow live in 30 days. Here is what you get and what you don't.

Abstract illustration of a magnifier connected to a process-flow diagram, clock, bar chart and gear, representing a 30-day AI operations assessment that maps processes, hours, cost and systems.

You are being asked to commit 30 days and a fixed fee before anything runs. That is a fair thing to push back on. This article walks through exactly what happens in an AI operations assessment, what you hold at the end, and what the assessment does not cover, so you can judge the value before you book.

Key Takeaways

  • An AI operations assessment is a fixed-scope, 30-day phase. It ends with a process, hour, cost and system map, one workflow running live, and a design for your first department.
  • You do not wait 30 days to see proof. One real workflow goes live inside the assessment, on your own systems and your own AI keys.
  • The assessment does not launch a full department. Loading brand knowledge, building the wider set of workflows and recording a performance baseline happen in the Launch phase.
  • Your main contribution is time with the people who do the work, not a technical project.
  • The map and the department design are yours to keep, whether or not you continue with Harnyss.

What an AI operations assessment is, and where it sits

AI tools answer prompts. People still copy, chase, check and file the work. For most companies the gap is not a shortage of AI tools. It is work that has never actually been handed off. The assessment is the phase where we find out which of that work should move, how much it costs you today, and what it takes to move it safely.

The managed engagement runs in four phases:

  1. First call. 30 minutes. You bring one time-consuming process. We review your current work, systems and model options.
  2. Assess. 30 days, fixed price. This is the phase this article covers.
  3. Launch. Usually 30 to 60 days. We load your brand and knowledge, connect systems, build workflows, set approvals and record a baseline.
  4. Run. Ongoing, or for a fixed term. Daily checks, exception handling and a monthly results review.

The assessment is the bridge between a conversation and a working department. It is short on purpose. It is also the phase where we are still learning your operation, so it is where most of the useful questions get asked.

30 days
Fixed-length Assess phase
1
Live workflow inside the assessment
4
Areas mapped: process, hours, cost, systems

What happens in the 30 days

The assessment has four working parts. They overlap, and the order depends on your business. We confirm the sequence with you on the first call.

1. The process map

We start with the work, not the technology. We sit with the people who do it and map how it moves: where it starts, who touches it, which system it lives in, where it waits and where it goes wrong.

You will usually see the same patterns. Orders arrive by email and get retyped into another system. A status update is chased by hand every week. A report is rebuilt from the same three sources each month. These are the repeated, weekly tasks that take at least one person's time, and they are where most teams start.

The map shows each process as it runs today, not as the handbook says it runs. That gap is often the first useful finding.

2. Hours and cost baselines

Next, we put numbers on the work. For each mapped process we record how many hours it takes and what that time costs. We work with the people who do the work, because they know where the hours go and a spreadsheet does not.

The baseline does two jobs. It tells you which processes are worth moving first. It also gives you a before picture, so that any later claim about results has something to be measured against.

Research on knowledge work points the same way. According to the Asana Anatomy of Work Global Index 2023, workers spend 58% of their day on what Asana calls "work about work": time spent coordinating work instead of doing the skilled, strategic jobs they were hired for. A baseline shows you what your share of that looks like in your own company, in hours and money, rather than as an industry average.

3. The systems review

An AI department is only as useful as the systems it can reach. In the systems review we look at where your work lives and how an agent could get to it: email, an API, a database or a web portal.

Three points matter here:

  • Nothing is migrated. Agents work in your existing accounts, through your APIs, databases or a browser.
  • Database access is read-only by default. Writes need explicit permission, and queries have row and time limits.
  • Gaps get named early. If a system has no workable way in, we say so now, not in month two.

The review also tells us which approval level fits each kind of work. Harnyss agents run in one of three modes: Requires approval, Notify only or Autonomous. Anything touching customers, contracts or money is a candidate for Requires approval. The mode is set per agent, and it is your call.

4. One live workflow

This is the part that separates an assessment from a slide deck. Before the 30 days end, one real workflow goes live.

It runs in your workspace, on your own AI provider keys, against real work. Every output gets a quality check, work that fails the check is held, and your team approves what matters from one inbox. You see what an agent-run process looks like in your operation, with your data, your tone and your exceptions.

We pick the workflow together, based on the process map and baselines. It is usually one that is repeated, well bounded and easy to measure. Which workflow qualifies depends on your systems and your risk tolerance. Confirm the specifics with the Harnyss team on the first call.

The live workflow is not a demo environment. It is a first real step in the department you may later run. If you stop after the assessment, the workspace, the workflow and your keys remain yours.

What you hold at the end

When the 30 days finish, you have concrete outputs, not a recommendation to "explore AI":

  • A process, hour, cost and system map of the work we reviewed.
  • One live workflow with a run history you can inspect, including quality scores and approvals.
  • A first department design. This sets out which workflows the department would run, which systems it connects to, where approvals sit, and who on your team does what.

The department design is the document you take into a decision. It lets you compare the managed path and the keys path with real scope in front of you. Harnyss proposals show both paths side by side.

What the assessment does not include

Clear limits are part of the value. The assessment does not include:

  • A full department launch. Loading your brand and knowledge, connecting every system, building the full set of workflows, setting approval steps across the department and recording a performance baseline belong to Launch.
  • Ongoing daily operation. Daily checks, exception handling and monthly reviews are the Run phase.
  • Migration of your systems. There is none, in any phase.
  • Replacing your people. The repetitive work moves to agents. Direction, judgment and approvals stay with your team.
  • A published price. The assessment is fixed price, but the amount is not published. Confirm it with the Harnyss team.

What we need from you

The assessment is light on your engineering team and heavy on your people's knowledge. We need:

  • Time with the staff who do the work. This is the main input. They explain how the work really runs.
  • A senior sponsor who wants it to succeed and can clear obstacles.
  • Access to the systems involved, through email, API, database or web portal.
  • A person to approve the live workflow's first outputs.

You are a good fit if you have repeated weekly work that takes at least one person's time, systems an agent can reach, a sponsor and an approver.

Is an assessment worth paying for and waiting on?

This is the honest objection, so here is a direct answer.

"I can work this out myself." You may be able to. The cost is your own time and the time of the people best placed to map the work, who are also the people running it. The assessment puts our operators next to your team so the mapping happens alongside daily work, not instead of it.

"I would rather see it work first." That is what the live workflow is for. You do not buy a plan and hope. Inside the 30 days you can inspect a real run, its quality score and its approval trail.

"What if it shows AI is not a fit?" Then the assessment has done its job at the lowest cost it can. The risk of a poor fit is real. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value or inadequate risk controls. A fixed-scope phase with a baseline and a working example is a way to test those three risks before you commit to a larger build.

"What if I decide not to continue?" You keep the map, the department design, the workspace and your keys. There is no lock-in.

How to decide

Ask yourself three questions:

  1. Can you name one process that eats at least one person's week, every week?
  2. Do you have someone who can give us time with the people who do it?
  3. Would you rather decide on a department with a live example and a cost baseline in hand?

If the answer to all three is yes, the assessment fits. You can read how the platform keeps agents governed on the platform page, check the security page for data handling, or see how the work looks in the Operations department. The work with us page sets out the managed path and the keys path side by side, and compare shows how Harnyss differs from automation tools and agencies.

FAQs

How long before I see something running?

One workflow goes live inside the 30 days. We choose it with you after the process map and baselines show which one is the best first step. Confirm timing for your situation on the first call.

Do I need a technical team?

No. You need people who know the work, a sponsor and an approver. We connect to your systems through email, APIs, databases or web portals that already exist.

Will the live workflow send anything without my approval?

Only if you allow it. Approval level is set per agent. Work that fails the quality check is held at every level.

Who owns what the assessment creates?

You do. The workspace, agents, workflows, skills, brand knowledge, AI provider keys and tool accounts are yours on every path.

What happens after the 30 days?

You choose. You can move into Launch on the managed path, where we run it for you, or on the keys path, where we build it and your owner takes it over. You can also stop. Both paths are priced in the proposal, side by side.

Next step

If you have one process you would like mapped, book an assessment →. It starts with a 30-minute call. Bring the process that takes the most time. We will tell you plainly whether it is a good candidate.

Sources