Six years ago I wrote a post about what Uber taught me about interviewing. The premise was simple: at a company growing from 7k to 16k people in eighteen months, hiring was the highest-use thing anyone could do, and most of us were doing it badly. The diagnosis held up. What changed is the tooling.
Back then, the bottleneck was data. Interviews were a black box. You could insist on structured questions, mandate detailed feedback, train interviewers, and still end up with hiring decisions that came down to who asked what, who was tired, and which recruiter happened to write the loudest summary. The whole apparatus rested on the assumption that you couldn't actually see what happened inside an interview room.
In 2026 that assumption is dead. Every interview can be captured, transcribed, scored against the rubric the hiring manager set, and aggregated into a dataset the recruiting org actually owns. The Uber-era problems I named in 2020 (lack of consistency, lack of transparency, dealing with uncertainty) are no longer infrastructure problems. They are operating problems. And the gap between teams that have solved the operating side and teams that haven't has become the real differentiator in how companies hire today.
What Uber actually taught me
I joined Uber in March 2016 and spent the next two-plus years on the hiring panel for nearly every Product Manager, Designer, Data Scientist, and Product Marketing Manager role in EMEA. The headcount went from roughly 7,000 to over 16,000 in that window. The volume forced me to interview constantly, and the variety forced me to think hard about what I was actually doing each time I sat down across from a candidate.
Two things from that period have only gotten more true. The first is that interviewing is not a distraction from your job. It is the job, especially for any leader trying to scale a function. Uber's engineering org treated "lead interviewer" status as something you earned through shadowing and apprenticeship, and tied it to personal objectives. Most companies, then and now, still talk about hiring as an extracurricular activity squeezed between real work. That framing is the original sin.
The second is that the quality of your interview process is bounded by your willingness to make it explicit. Dividing assessment areas across interviewers, mandating detailed written feedback within fifteen minutes of the conversation, and pushing back on lazy "no-hire" sentences without evidence all work because they force the panel to be specific. Specificity is what lets you compare candidates and lets you coach interviewers when they drift.
The best interviewers get into the habit of committing fifteen to twenty minutes after the interview to expand on their written notes and provide proper feedback for all candidates. Do this. Every time.”
The three problems that never went away
When I wrote the original post in 2020, I called out three things that even Uber, the fastest-growing company of its era, had not solved. They are the same three things almost every recruiting team is still wrestling with in 2026.
Lack of consistency. From a candidate's perspective, the outcome of any given interview has a surprisingly large arbitrary component. Which questions get asked, who is on the panel, what kind of day that interviewer is having, whose role the hire reports into. Companies tried to fix this with structured interviews (a fixed set of questions every candidate gets). It helped, but it also flattened the conversation and killed the ability to chase an interesting thread. The trade-off was real and most teams accepted it as the cost of scale.
Lack of transparency. Almost no company outside the recruiting function carries hiring quality as a personal objective. The CEO will tell you hiring is the most important thing she does, and her direct reports will agree, and then none of them will have a single OKR tied to interview quality, time-to-decision, or hire performance at 90 days. The opaqueness of what actually happens inside an interview lets bad habits compound. Interviewers don't prepare, they talk too much, they confuse rapport for assessment. Training programs exist but their impact is unmeasured.
Dealing with uncertainty. Some hires are easy calls, most are not. The candidates in the middle of the distribution are where the calibre of your team actually gets decided, and they are also where panels have the least information. You either err on the side of caution and miss out on real talent, err on the side of risk and absorb the cost of bad hires, or invest in another interview round and damage the candidate experience while extending time-to-hire. Every choice has a downside, and the choice usually gets made with thin evidence.
The reality is that without being able to measure what is going on in interviews, they are hard to manage at scale. And without engendering a repeatable and consistent interview process, hiring decisions will remain far too arbitrary.”
What AI actually changes in the interview room
The thing I could not have predicted in 2020 is how completely the data problem would dissolve. Every interview can now be captured live, transcribed accurately, structured against the rubric the hiring manager defined, and rolled into a dataset that lets a TA leader actually see what is happening across the function. The black hole I described back then is gone. What remains is the question of whether teams use the data well.
The instinct most teams have when they first turn on capture is to use it as a backstop. They review the transcript when a decision feels wrong, or when a candidate complains, or when a hiring manager wants to second-guess the panel. That is a reasonable starting place, but it underplays what is possible. The real value shows up when you use the captured signal to coach interviewers in real time, to standardize the rubric across panels, and to feed quality-of-hire data back into the calibration process.
Two specific moves matter more than anything else. The first is making the rubric the single source of truth, not the interviewer's gut. The second is closing the loop between the rubric and what actually happens in the room. When you have both, structured interviews stop being a trade-off (consistency at the cost of conversation) and start being a floor everyone clears. The conversation can be as fluid as it wants because the assessment is being captured underneath it.
Capturing interviews without a clear rubric just produces more unstructured data. The tooling makes the rubric the use point, not the transcript.
The trust gap between recruiters and hiring managers
The Uber-era diagnosis missed one thing, and it is the thing that surfaces most when you actually have visibility into the interview process. Recruiters and hiring managers do not agree as often as either side likes to think. The misalignment is usually invisible because nobody has the data to argue with, but once the conversations are captured and the rubrics are explicit, the disagreement becomes legible.
According to Metaview's 2026 AI & Hiring Alignment Report, surveying 505 recruiting leaders and hiring managers across North America and EMEA, 58% of recruiters say they regularly disagree with hiring managers about candidate quality. The gap shows up in everything downstream: time-to-hire, hire performance at 90 days, candidate experience, the recruiter's confidence in pushing back on weak feedback. It is not a personality problem. It is a structural problem caused by the absence of a shared evidence base.
The teams that close this gap fastest are the ones who treat the interview record (the rubric, the captured conversation, the structured feedback) as the shared artifact both sides argue against. When the conversation is "the candidate's response to the technical question doesn't satisfy criterion three" instead of "I just don't think they're senior enough," the disagreement becomes productive. The recruiter and the hiring manager are no longer arguing about taste. They are arguing about evidence.
Old heuristics vs new heuristics
What I was actually doing at Uber, and what most thoughtful interviewers were doing in that era, was running a set of compensating mental heuristics to deal with the lack of data. Some of those heuristics still hold up. Others have been quietly replaced by what the tooling can now do.
- Mandate detailed written feedback to force thoughtful interviews.
- Use structured questions to fight bias, accept the conversational cost.
- Train interviewers and hope the training sticks.
- Panel debriefs as the moment where evidence gets aggregated.
- Capture every interview and let the rubric drive the structured feedback.
- Let the conversation flow; the assessment runs underneath it against criteria.
- Coach interviewers from their actual transcripts, not abstract best practices.
- Panel debriefs as a check on the structured record, not a re-derivation of it.
Where AI gives recruiting teams use
Look at how a recruiting team's day actually breaks down and you can map the AI use points cleanly. Sourcing eats time before the interview. Application review eats time on the way in. The interview itself was historically the part nobody could see into. Reports are the layer that closes the loop. Each of those pieces now has a piece of tooling that handles the mechanical work and frees the recruiter or hiring manager to spend their time on judgement.
The Uber answer was a giant recruiting function. The 2026 answer is a sourcing agent that builds the pipeline against the rubric the hiring manager actually wrote.
Most "no-hire" calls in 2020 came down to triage capacity. AI app review puts every applicant against the same ICP so the top of funnel stops getting decided by who reviewed the resume.
The black box I described in the original post is now lit up. Live capture and structured notes against the rubric replace the "fifteen minutes of writing after the interview" ritual I used to preach.
OKRs for hiring quality were impossible in 2020 because nobody had the data. Reports turn the captured interview record into the dashboard that makes hiring a coachable function.
The point is not that you adopt all four tomorrow. The point is that for the first time the four core problems of an interview process (volume, signal, judgement, learning) each have a piece of infrastructure that can absorb the mechanical work. The recruiter and hiring manager get more time on the part that actually requires a human.
The headline numbers from the AI & Hiring Alignment Report give a sense of the prize sitting on the table for teams that get the operating model right.
None of these numbers solve the problem on their own. They tell you where the use is. The teams that pay attention are the ones who stop treating hiring as a craft each recruiter practices alone and start treating it as a system the function operates together.
The operating shift
If I were rejoining Uber today and the CEO asked me what would actually change hiring outcomes at their scale, here is what I would say. It is shorter than the 2020 list because the tooling now does the heavy lifting on the parts I used to have to evangelize.
One: write the rubric first. Before the job description, before the sourcing strategy, before the first interview, the hiring manager and the recruiter agree on what good looks like in concrete criteria. The interview process exists to gather evidence against those criteria, not to surface opinions.
Two: capture every interview against that rubric. Not for surveillance. For evidence. The transcript and the structured assessment become the shared artifact the panel argues against. The bad habits I called out in 2020 (interviewers who don't prepare, talk too much, confuse rapport for assessment) become visible and coachable for the first time.
Three: measure hiring quality the same way you measure any other function. The reason OKRs for hiring did not exist outside the recruiting team in 2020 was that nobody had the data. The data now exists. The reason it should exist is that hiring is the thing that determines whether everything else works. The difference between a good interviewer and a bad one compounds on every hire that follows.
Four: close the loop on quality of hire. Six months after a hire starts, what does the rubric they were assessed against say about how they're performing? This is the feedback signal that turns hiring from a one-off transaction into a function that learns. The sourcing layer gets sharper, the application review gets more accurate, the interviewer coaching gets more specific. The whole stack improves because the inputs and outputs are connected.
Bring Metaview into your hiring stack.
Live notes, structured scorecards, and ATS sync - set up in under 10 minutes.
Frequently asked questions
Is structured interviewing still worth it in 2026?
Yes, but the trade-off has changed. The old reason structured interviews felt restrictive was that the rubric was the only mechanism keeping the conversation honest. With interview capture, the rubric runs underneath the conversation as a check, so the interviewer can have a more natural exchange and still produce structured evidence. Treat the rubric as a floor, not a ceiling.
Why do recruiters and hiring managers disagree so often?
Mostly because they are operating against different mental rubrics that were never written down. The recruiter is screening for "would this candidate pass the bar" and the hiring manager is screening for "would I personally hire this candidate for this specific role." Those are different questions. When the rubric becomes explicit and the interview evidence is shared, the disagreement becomes evidence-based rather than taste-based, and the conversation gets productive faster.
Does interview capture hurt the candidate experience?
Done badly, yes. Done well, no, and arguably it improves. Candidates told upfront that the conversation is being captured for note-taking and scorecard purposes generally appreciate that the interviewer is paying attention to them rather than typing. The candidate-facing pitch is that the captured interview means the hiring decision is being made against the same evidence everyone on the panel sees, which is a fairer process than the alternative.
How do you handle the middle of the candidate distribution?
This was the hardest problem to solve in 2020 and it remains the highest-use place to focus. The shift is that you have more evidence on those middle candidates than you used to. Instead of choosing between caution, risk, or an additional interview round, you can pull up the structured assessment against the rubric, see exactly which criteria are weak, and decide whether to schedule a focused follow-up on those criteria or pass. The decision becomes targeted instead of binary.
Should hiring be in every leader's OKRs, not just the recruiting team's?
Yes. The reason it usually isn't comes down to whether hiring quality is measurable, and historically it wasn't measurable outside the recruiting function. That excuse no longer holds. With structured interview capture and quality-of-hire reporting in the same system, you can build a hiring-quality OKR for every leader who participates in the panel. The signal you want is consistency of assessment against the rubric and 90-day performance of the hires they greenlit.
