Welcome to episode ten of Pinpoint's How-To Series, where we discuss how to avoid the 3 biggest mistakes TA teams are making with AI
Right now, most TA teams fall into one of three camps when it comes to AI.
The first aren't using it at all. Usually not by choice. Legal, IT, or procurement has said no or not yet, and nobody can quite work out how to change their mind. Meanwhile, they're watching the same headlines as everyone else and wondering if they're falling behind.
The second are using it cautiously. They've switched on a feature or two, but they worry about bias, about getting something wrong in front of candidates, and about what happens if a hiring decision gets challenged. So AI stays at the edges, often used in disconnected ways, and never trusted with anything that really matters.
The third have raced ahead. The early adopters who've built clever workflows and processes on their own. It's impressive to watch, but hard to replicate across a whole team.
What's striking is that none of those positions are wrong. But very few teams are getting what they hoped for from AI.
That's rarely a technology problem. It's usually a foundation problem.
For episode 10 of the Pinpoint How-To Series, I spoke with Matt Alder, host of the Recruiting Future podcast, and Jessie Schofer, founder of Stakkd. Here's what came out of that conversation: three problems they see coming up again and again, and what good looks like instead.
The most common starting point for AI in TA isn't a problem to solve. It's a mandate from above.
Leaders tell teams to "use AI," but without being clear on what that means or what success looks like. Teams feel the pressure to do something, so they experiment, often without structure, without measures, and without any way of knowing whether it's working.
The results can be counterintuitive. Some teams see productivity drop during the experimentation phase because people are so engaged with the new tools that they let their actual work slide. Others run pilots for a few weeks, see nothing dramatic happen, and quietly shelve the whole thing. Nothing happened because nobody made it happen.
Part of the problem is that AI is genuinely hard to orientate around. The technology moves faster than most organizations can absorb, and it's uneven in what it can and can't do. That makes it difficult to know upfront whether AI can even solve your specific problem, which is one reason people reach for pilots without a clear outcome in mind.
Start with the problem, not the tool. The teams that get value from AI fastest are usually the ones that arrive with a very specific thing they want to fix. Not "transform hiring with AI," but something like: we're losing candidates at the offer stage and don't understand why, or our time-to-fill for technical roles is two weeks longer than it should be.
Pick something you can measure. If you can't tell whether it worked, it's the wrong starting point.
A useful frame: start at the individual level, then bring it to the team. If you haven't been given clarity from leadership, create it yourself. Work out where AI might help you personally, share that with the team, and build a conversation from there. Get a whiteboard up. Run an ideation session. You can usually land on two or three places where the effort is real, the business need is clear, and AI could plausibly help. From there, you have enough to run a focused experiment with something to measure against.
The goal has to come before the tool.
When hiring data lives across multiple systems, spreadsheets, email threads, and individual inboxes, AI only ever sees part of the picture. The output reflects that.
It's not just systems, either. A lot of hiring knowledge lives in people's heads. The person who knows how the end-to-end process actually works. The handoff that only exists because two colleagues figured it out between themselves. That tacit knowledge is invisible to AI, and invisible to anyone who needs to work consistently across a team.
Then there's the sprawl problem. Right now, almost every tool in the market is building its own AI layer. That creates a real question about consistency: what is our AI, what is it doing, where does candidate data go, and are these tools working together or just adding to the noise? Without a strategy for how they fit, you end up with duplication, compliance risk, and a candidate experience that feels disjointed, because it is.
Context is everything for AI quality. The more context it has about your business, your process, and your candidates, the better the output. Scatter that context across a dozen tools and nobody benefits.
Start by mapping your hiring process end to end. Understand where your data actually lives: which systems hold what, how information moves between them (or doesn't), and where the gaps are. Most teams haven't done this explicitly, and it's hard to fix what you can't see.
Then take stock of what your existing tools can already do. Before adding anything new, understand what's already available in your current stack. Most teams significantly underestimate it.
From there, the goal is consolidation. The more of your hiring that lives in one place (job data, candidate information, interview notes, feedback, context from earlier in the process), the more useful AI becomes. It stops working with half the picture. The team gets faster, more consistent, and less reliant on the one person who knows where everything is. Candidate experience improves too, because the process reflects a complete view of that person rather than fragments pulled from different tools.
You don't have to get there all at once. But the direction of travel matters: fewer systems, better connected, more of your hiring in one place.
The question of where AI ends and humans begin doesn't have a fixed answer. That's part of what makes it hard.
The line shifts as AI capability grows. It shifts with organizational risk appetite. It shifts with jurisdiction, regulation, and employer brand. What's right for a frontline retailer hiring hundreds of people a week is different from what's right for a professional services firm hiring for senior roles.
Regulatory clarity isn't helping. Many of the legal questions around AI in hiring remain unresolved. Cases are mid-hearing. Regulations are being interpreted. Guidance is still being drafted. That uncertainty makes organizations rightly more cautious. But caution that becomes paralysis is its own mistake, and it doesn't protect you from the more immediate risks.
The practical risks right now are often simpler than the regulatory ones. People entering candidate data into open general AI models because they don't know they shouldn't. Teams making AI decisions without looping in IT or legal, not out of recklessness, but because those conversations haven't happened. Candidates who are increasingly aware that AI is part of the process but have low trust in how it's used, often not based on what's actually happening, but on what they don't know.
Decide deliberately where humans make the call. Don't leave it to chance or assume everyone on the team has the same view. Map it out. Where does human judgment add something AI can't replicate? Where is a human check a legal or ethical necessity? Where does it matter for the candidate experience? Put that in writing, as part of your process design.
Invest in AI literacy across the team. Organizations that mandate AI without supporting the team to understand it are setting people up to make mistakes. Upskilling isn't optional, and it's not purely a leadership responsibility. Individuals should be asking questions, seeking clarity, and not assuming they know something they're not sure of.
Be transparent with candidates. They deserve to know how AI is involved in the process. That transparency is both a legal expectation in many jurisdictions and a practical trust requirement. Teams that are clear and open about their AI use tend to attract better candidates and avoid the reputational risk that comes with opacity.
Ask your vendors the hard questions. Explainability and transparency aren't nice-to-haves. They're what IT and legal will ask for, and they're what candidates increasingly expect. Understand how the tools you use arrive at their outputs, and push back when you can't get a clear answer.
Solutions Engineer at Pinpoint
