Your AI tool isn't broken, but your process is


For the past couple of years, everyone has been racing to buy the latest AI tools. Meanwhile, each vendor is pushing promises, pilots, discounts, and multi-year contracts. The AI buying frenzy has created a lot of pressure to find the next gamechanger. Every vendor insists that their product is the best on the market.
It’s exciting and fun to learn about all the new tech, but it’s also hard to navigate. Every vendor is pushing a pilot to try out their new and exciting tool, and while we could jump in and test out a bunch of new products, that’s just not the best way to operate.
There is an important question we need answered first.
What problem does the shiny new AI tool actually fix?
What process is broken? How will this tool fit into my workflow? Will it fix a bottleneck or just push it further down the line? Will we just be creating a new problem to solve?
The real challenge isn’t choosing the right tool, it’s understanding your process. Otherwise, you’re not really transforming your workflow. You’re just bolting AI on top and hoping for the best. And even when the technology works exactly as intended, that doesn’t mean it’s going to deliver the results you want. If people don’t understand it, trust it, or change how they work to incorporate it, a technically successful AI tool can still fail.
That’s the real issue to me. To stay ahead, we have to be up to date on the latest and greatest technology, but when we’re moving so fast, we can miss the big picture.
Experience learned the hard way
At Horizontal, we’ve made this mistake before.
We experienced it during our early days of implementing our VoiceAI agent, Alex. We had thousands of applicants who needed to be screened, so we thought, let’s put Alex to work.
From the start, we didn’t take that step back to think strategically. We just assumed the problem was recruiters didn’t have enough time to review applications, so implementing AI would make things easier by screening candidates as they applied.
To be clear, it solved our problem. It did a great job of reaching out to all of those candidates and conducting a short screening call.
The issue was with how we went about it.
This new shiny tool just moved the bottleneck down the line. Suddenly recruiters had hundreds of screened candidates to review and follow up with. We hadn’t helped save them time, we’d just moved it to another step in our process.
That’s because we were fixing the wrong problem: It wasn’t the volume of untouched applications. The challenge was finding the right candidates efficiently.
So we had to take a step back and look at the bigger picture. What was the actual problem we were trying to solve and how could we rework our workflow to solve for it?
So, that’s what we did.
Instead of asking Alex to conduct screening interviews with hundreds of candidates, many lacking any relevant professional experience, we shifted our focus to identifying and prioritizing qualified candidates earlier in the process. That way, recruiters could spend more time engaging with people who were a stronger fit for the role.
Rather than asking AI to do more work, we needed it to help our recruiters do the right work.
That shift also meant rethinking how our team interacted with the technology, where human review was needed, and how we could use recruiter feedback to refine and improve the entire process.
AI won’t fix a broken process
That lesson still sticks with me. From the product side, it’s easy to get excited. It would not be hard for me to introduce a bunch of new and exciting AI tools that I see adding value from the product side.
But first, I need to understand what manual processes we’re trying to free our team from at the desk level. I need to know if the tool would actually save time and benefit our end users. I need to ask whether this will even work at-scale and how our overall workflow is going to change because of it. We also need to understand what the ROI would be, if we free up X amount of time, what are we filling that with? How does that translate to more submissions, more placements?
Ultimately, the goal isn’t AI adoption. The goal is better business outcomes. Sometimes AI is the answer. Sometimes redesigning the process is the answer. And often, the best result requires both.
A mature AI strategy isn’t just identifying where AI fits. It is also knowing where human judgment, relationships, sensitive decisions, privacy, or risk mean AI should assist rather than own the process.
In recruiting, that’s especially important. AI can help us identify potential matches, organize information, and connect with candidates faster. But evaluating people, building relationships, and making hiring decisions still require human judgment and accountability. The technology should support those decisions, not replace the people responsible for making them.
Our competitive advantage
I love partnering with our vendors. I love learning and exploring the new technology that’s hitting the market.
It’s all very exciting, but it also creates pressure to adopt technology simply because it’s the new shiny object.
Before adopting a tool, we should also be asking:
- Can it integrate with our existing stack?
- Who owns it?
- How will data be handled?
- Can we govern it?
- Can it scale globally?
Otherwise, we risk adding more technology without actually simplifying the process.
The organizations that gain a competitive advantage won’t be the ones buying the most AI tools. They’ll be the ones asking the right questions first.
At Horizontal, we’re not trying to put AI everywhere. We’re trying to put it in the right places, where it removes friction, improves outcomes, and gives our people more time for the work humans do best.
Connect with Kim Ullstrup, Product Operations Manager, on LinkedIn.