TL;DR

A faster task doesn’t automatically make the business faster. Follow the whole job, count the work that remains, and decide how you’ll use any capacity returned to the team. That’s how you find out whether the investment helped.

Before you approve another AI investment, sit with the person who’ll use it and ask what they expect to stop doing.

Give them time to answer. They may tell you about entering vendor quotations, finding documents, checking whether purchasing received a reply, or updating a spreadsheet so somebody else knows where a repair stands. They’ll probably describe the work more clearly than the proposal does, because they’re the ones doing it every day.

That’s where I think the AI conversation needs to start. The technology can do extraordinary things, but the business still needs to understand which problem it’s solving and what happens to the people responsible for the work afterward.

We’ve managed big technology changes before

Around 2014 or 2015, at my previous company, we decided to move from on-premise systems to the cloud. I remember how uncomfortable that felt. There was plenty to be excited about, but we were changing something the company depended on, and we still had to keep the business running through it.

We asked questions about what would change, who would be responsible, and what we expected to gain. Those were normal questions. Even in 2021, nobody would have found it strange to ask how a technology investment would improve the operation.

Then generative AI arrived, and suddenly you could watch an hour’s work happen in minutes. Of course people started imagining what that could mean across an entire company. I did too. But a faster task can still feed into the same queue, wait for the same approval, or depend on information somebody hasn’t updated.

The person got faster. The customer may still be waiting.

The work left behind belongs in the calculation

Consider a proposed workflow connecting your inboxes, spreadsheets, document storage, and ERP. It reads incoming information, moves it between systems, and gives the team a polished interface. There may be real value in that. I’d still want to understand who maintains the information underneath it and what happens when something changes.

One operator described her situation this way:

Someone created those tabs to keep track of something the operation needed. Before replacing or connecting them, find out what they carry. Otherwise, you can spend considerable money changing where people look while leaving them responsible for the same checking and reconciliation.

The invoice won’t tell you how often somebody had to correct a part number, investigate a missing attachment, or explain why two systems showed different answers. Ask the team. That work belongs in the investment calculation too.

A customer should be able to recognize the improvement

Before one customer signed up, the conversation came down to a seventy-five-part request and manually entering the vendor quotations coming back. Read the reply, identify the part, enter the quantity, condition, price, and lead time, then associate the supporting documents and move to the next line.

As that grows toward thirty, forty, or fifty quotations, someone’s day starts disappearing into entry before the commercial decisions are finished. For that customer, iRFQ was the final push toward signing up. He could recognize the work it addressed.

That gives us a useful place to evaluate the improvement. Which inputs can be handled? What still requires review? How much correction is involved? Does the information reach the person preparing the customer’s response?

If you run repairs, follow the same question from the work order. A technician identifies a material requirement, purchasing starts sourcing, vendor replies arrive, and somebody needs to establish whether the job can move. Improving one step helps, but the repair still waits if the next person doesn’t have what they need.

Company knowledge should help the next person

Getting information into the operation is only part of the opportunity. Your business already has experience worth using: supplier history, previous transactions, customer requirements, and the reasons people made particular decisions.

That’s part of the thinking behind ELIA. We spent almost a year exploring and developing it before alpha, initially deploying it to serve as centralized knowledge. The larger opportunity is helping people use that knowledge while they’re making decisions inside the operation.

A customer explained the outcome he wanted from keeping documents with transactions:

In that same conversation, he wanted a colleague to find answers “without always coming to me.” Most operators know exactly what that means. There’s somebody who remembers the history and gets interrupted whenever the rest of the team needs it.

Their experience should make the company stronger. More of it should be available without requiring another call.

Research offers a useful example. A study using data from 5,172 customer-support agents found a 15% average increase in issues resolved per hour after AI assistance was introduced, with less experienced employees benefiting most. The researchers found evidence that the assistance helped distribute practices associated with stronger performers. That’s a result from a particular setting, but it shows the value of making organizational experience useful during the work. Generative AI at Work​

This is the conversation driving back the value to real, tangible output.

Prove it on one workflow

Before expanding, pick work your team can explain from beginning to end. Establish how much active time it takes, where it waits, and how often somebody has to correct something. Then follow the assisted process through to completion, including review and exceptions.

If entry gets faster but checking consumes the saving, you need to know. If the team handles more demand and maintains quality, that’s worth understanding too.

Then decide where the recovered capacity goes. More selling, better sourcing, faster customer responses, or room to grow without adding the same administrative burden? Time returned to the team needs a purpose.

That’s the conversation I want to have with operators. Show me the work, show me where your people get stuck, and make us explain what changes. Before we talk about how much more the business can accomplish, we should be able to tell the people doing it what they finally get to stop doing.

Thanks for reading. This work was built on my recent LinkedIn article. If this reminded you of something your team deals with every day, I’d like to hear about it. Show us where the work gets stuck, and let’s talk through what could make it easier.