An AI project can hit every productivity KPI without improving EBITDA. Saving employees time doesn’t automatically reduce costs or grow revenue. Our clients are smart. They generally know where the big buckets of investment opportunity are within a business. But if there are ten areas, it’s tough to determine which one to tackle first.

Which opportunities will actually move the P&L? That’s the question.

And that’s the work we own at AIVC, from discovery through execution. We work with frontline teams and management to connect changes in day-to-day operations to financial results. We need to understand the actual processes today, how we want them to be different, and what needs to change to get from point A to point B.

Measurement is the spine of that work. Without clear attribution—what we’re doing impacts this, this unlocks capacity, that capacity changes revenue or costs—there’s no way to know whether a project is delivering its intended outcome. Once you have that connection, you can make better decisions about what to invest in next.

Start with the business decision

One of our clients had already identified call centers as an area worth looking at. Their objective in the near term was immediate cost savings, with a lower risk profile. Our recommendation was to identify areas where we could deploy vendors that had proven themselves in the market.

That’s starting with the business: what is the objective, to what degree, and what’s our timeline? Any decision we’re making from a process, technology or people perspective is tied back to those anchors.

Any time you’re hiring somebody or building a new project, that’s an investment. You want to answer two questions from the finance lens. One is: am I going to generate positive ROI? Am I going to make more than I spend?

The second is: of all my projects with a positive ROI, which one generates the highest return in the time period that’s important to me? That could be a year, an entire investment holding period, or until the next board meeting.

Measure every step

The client’s goal was to free capacity from the 90 call center FTEs. To measure that target, we found that “average calls per person per day” was the right metric to track. There are two factors here. One is time spent on the phone. The other is the work after the phone call, entering a name, birthday and other information.

If you just look at time on the phone, that’s pretty short. Looking at calls per day includes the work we’re trying to take away afterward, thereby impacting the full workflow, not just a single step.

If a call hit a roadblock and got forwarded to an agent, that didn’t count as a fully handled call. Taking on 50% of the business outcome didn’t achieve the outcome we were trying to deliver.

We sat with the teams operating the call centers and the operational folks to map out their current process. Every clinical provider had their own preferred workflow. Every location had a little bit of a different way of doing things. 

Measurement goes beyond accuracy or throughput rate: can we free measurable human capacity, or generate additional revenue with our current resources? For the call center, we needed to know how much work could come out before we could estimate how spending might change.

Make the pilot answer the financial question

We identified two vendors and piloted each at one location. What people tell you in a sales demo may or may not reflect reality. Things work differently in different environments. You prove something before you scale it.

Anything in the original business model was going to be benchmarks and case studies the vendors provided. The pilots gave us data from the company’s own environment, operations and call centers to feed back into that model. If we did scale this, what could the total potential outcome be? That was a projection to test, not savings already realized at scale.

The metrics we care about are typically pretty different from what a vendor cares about. A vendor is going to show you things that make them look good, and they might define things differently than we do.  Also, by allowing vendors to own metrics, they’re grading their own homework, on often the wrong metrics.  

On the call-center project, we had to reconcile the vendors’ outputs each week into a structure that aligned with how the company thought about cost reduction. We needed the metrics to support the business, focusing on freeing up capacity to do more and better work, rather than simply cutting costs.

There’s a gap between what people on the ground are doing and how that tracks back into the financial statements. Understanding which levers to pull at different layers of the organization is a big part of the work.

We work with the CFO to map the actual logic. We have working sessions to align on that logic, then weekly readouts: has anything materially changed based on new learnings or new data?

Once we hit this target, what spending can change? Or how much additional capacity does it unlock? You have to understand those connections to actualize the value.

Use the evidence to choose what comes next

The call-center pilot gave us evidence to put back into the business model, ultimately unlocking a 5% EBITDA lift. That connection between the work and the investment decision is what we want to carry into every project.

When you understand what’s performing well and what isn’t, you can reprioritize what you’re investing in. Are we using too many resources or too few? What are the winners? Let’s get those winners more money and take money away from things that aren’t giving us value.

The question isn’t just the total potential. Can we measure value in smaller increments than a multiyear period?

That’s the larger ambition. With the measurement layer set up, we know how to size a new piece of work. We know what its acceptance criteria mean and how they translate back into the financial statements. We can understand which opportunities are worth our time, and use the value we generate to fund more of them. 

You can’t invest in what you can’t measure – that’s the heart of our work.