Takeaways from my conversation with hiring leaders at Tally
Anyone who has spent real time in hiring knows the partnership between hiring managers and talent acquisition teams is the thing that quietly decides whether a process produces great hires or a string of near-misses. Most of the conversation in our industry focuses on funnels, sourcing tools, and rubrics. The cross-functional muscle between Eng and TA gets talked about a lot less, and it shows.
That is why we kicked off a new event series at Metaview that examines the cross-functional team sport that is hiring engineers. The first conversation was with Jan Chong, VP of Engineering at Tally and former Senior Director of Engineering at Twitter, and Beryl Wang, Senior Technical Recruiter at Tally. We spent an hour on how their team has built a consistent, high-quality hiring machine and what they do to keep the Eng+TA partnership tight.
What follows is my recap. Not a transcript, not a polished case study. Just the takeaways I keep coming back to from that conversation, and how I would translate them into operating habits for any team trying to raise the bar on hire quality without grinding their interviewers down.
Interviewing is part of the job, not a favor
The first thing that struck me in the conversation was how plainly Jan and Beryl talked about prioritization. Most companies treat interviewing as something engineers do on top of their real work. At Tally it is the real work, on equal footing with shipping. Without that framing, Beryl said, any hiring process will quietly degrade because the load falls on whoever feels guilty enough to take it.
What made it concrete for me was how Beryl described her dependency on Engineering leadership. She is not lobbying managers in DMs to free up an engineer for a panel. She is sending a regular signal of who is heavy in the loop right now, and Eng leadership factors that into roadmap planning the same way they factor in on-call rotations. Interviewing load is a planning input, not a planning afterthought.
Jan went further. At Twitter she added interviewing to the career ladder. To level up, you had to be making meaningful contributions to the hiring process. That is the kind of move that sounds small and is actually load-bearing: it removes the question of whether a senior engineer should care, and replaces it with a question of how well they are doing the job.
If you are at a healthy, growing company, you are going to be spending a ton of time interviewing and engaging in the hiring process. Like code reviewing, it is part of the job.”
Align on day-one skills vs. trainable skills
The second move I want to steal from Tally is the way Jan and Beryl scope a role. They do not ask "what would the perfect candidate look like" because that question has no useful answer. They ask what does this person have to walk in already knowing, and what can we teach them in the first 90 days. That single split changes the rubric, the questions, and the calibration conversations.
Jan calls a dedicated kickoff with the full panel, walks through the list of competencies that could conceivably matter, and then forces the team to cut the list down to five and rank them. The ranking is the part most teams skip, and it is the part that makes the rubric useful. When tradeoffs come up later (and they always do), the panel has a pre-agreed answer instead of an in-the-moment debate that drifts into whoever has the strongest opinion.
Beryl pointed out something honest that I think a lot of recruiters feel and rarely say out loud. The first version of the rubric is almost never right. The panel talks to a few candidates and realizes the thing they thought they cared most about is not actually the thing. Treat the first 3-5 candidates as rubric calibration, not just as candidates. Adjust the ranking, and tell the panel you adjusted it.
The day-one vs. trainable split is where this gets tactical. Tally uses Scala. They are happy to interview engineers in other languages and teach Scala on the job. They are not happy to interview engineers who do not communicate well and teach communication on the job, because that is much harder to teach. The pattern is: specify the things that are expensive to learn on the job, give yourself room on the things that are cheap to learn on the job. It is the same shape as a build vs. buy decision.
Do not inherit the process that hired you
This was the one that landed hardest for me. Jan put it plainly: most teams do not design their hiring process, they inherit it from whatever process happened to be running when they got hired. If you have not actively thought about your loop, you are almost certainly running a process built for a different company, a different role, and a different stage.
The cost of inheriting shows up in small ways that add up. Interviewers improvise questions because no one prepared a question bank. Two candidates get asked different things and the panel cannot meaningfully compare them. Feedback notes are thin because the interviewer did not know what they were supposed to be assessing. Bias creeps in not because anyone is acting in bad faith but because structure is the only thing that holds bias back at scale.
- Sourcing enough qualified candidates into the top of the funnel
- Getting hiring managers to spend any time on a panel at all
- Persuading leadership that hiring is worth investing in
- A scarcity of resumes that matched a job description
- Designing a loop that produces honest, consistent signal
- Calibrating panels so two interviewers grade the same way
- Onboarding new interviewers fast enough to keep loops moving
- Filtering signal from noise in an over-applied candidate pool
Tally invests in the upfront design work most teams skip. They build rubrics. They write question banks. They use Metaview to onboard new interviewers through virtual shadowing of hand-picked recordings, and they use it for ongoing calibration so the bar does not drift. The thing I want to underline: this is not glamorous work, and the payoff shows up six months later in retention numbers and hire quality, which is exactly why most teams never get around to it.
What the Eng and TA partnership actually looks like at Tally
One of the things I wanted to understand was the operating cadence between Jan and Beryl. What does a healthy Eng+TA partnership look like, day to day. The answer was less about ceremonies and more about shared ownership of the same metrics. They look at the same dashboards. They debrief loops together. When something breaks (a candidate drops, a panel misaligns, an offer goes sideways), they treat it as a joint postmortem, not a TA problem or an Eng problem.
The kickoff meeting Jan described for new roles is the closest thing to a ritual. Full panel in the room. Five competencies, ranked. Each interviewer leaves with a specific question area and the rubric for grading it. The whole exercise takes an hour and saves weeks, because every panel debate after that has a pre-agreed reference point.
Beryl's role in the partnership is not what I would call traditional sourcing-and-scheduling. She is closer to a chief of staff for the hiring loop. She runs the kickoff. She maintains the rubric. She gives Eng leadership a clear picture of who is loaded with interviews this week, and she pushes back when a panel skips a step. The skill set looks more like operations than recruiting, and I think that is the direction the senior TA role is heading at every serious company.
It takes a couple of tries to align on what skills a hiring team wants to prioritize for a particular role. The panel might speak to a few candidates and then realize that maybe what they initially thought they wanted is not what they actually care most about.”
Where AI gives recruiting teams use
The Tally conversation pre-dates a lot of the AI-in-hiring wave, but the conclusion I drew listening to them is that the teams who built the Tally-style operating habits in 2022 are the ones who benefit most from AI now. Structured loops, ranked rubrics, and recorded interviews give an AI layer something useful to do. Without that scaffolding, AI in hiring is just a vibe-check accelerator.
At Metaview we build the AI layer that sits on top of that scaffolding. AI Sourcing finds candidates against an Ideal Candidate Profile a panel has actually agreed on. Application Review ranks inbound applicants against the rubric the kickoff produced. Notetaker captures interviews so the calibration work Tally does manually can compound across every loop. Reports closes the loop by showing leadership which signals actually predict hire quality.
Run searches against the ranked rubric the panel actually agreed on, not the wishlist the hiring manager wrote alone.
Score inbound applicants against the day-one skills, separate from the nice-to-haves, so the panel sees only the candidates that match the bar.
Capture every interview so calibration, onboarding, and bias checks compound across the panel instead of evaporating after the loop.
Surface which rubric signals predict hire quality 90 days in, so the next kickoff is smarter than the last one.
The pattern in the data backs this up. According to Metaview's 2026 AI & Hiring Alignment Report, surveying 505 recruiting leaders and hiring managers across North America and EMEA, the depth of AI adoption (not the mere presence of it) is what predicts a healthy Eng+TA relationship. Teams that have AI woven into the core of their hiring workflow rate the partnership excellent at different rates than teams who use it occasionally.
The takeaway I would draw, looking at the Tally conversation through this lens: AI does not fix a broken partnership, and it does not replace the kickoff or the ranked rubric. It compounds the work teams like Tally were already doing. The question is whether your team has the operating habits worth compounding in the first place.
The operating shift for any team that wants this
If you want to move your team from inherited-process to Tally-style design, here is the sequence that I would run.
One: put interviewing on the career ladder. Make meaningful contribution to hiring a requirement for promotion to senior IC and above. This single change shifts the conversation from "can you spare an engineer this week" to "of course, this is part of the job."
Two: run a real kickoff for every role. Full panel in the room, list of competencies, cut to five, ranked. Each interviewer leaves with a question area. The kickoff itself takes an hour and saves the next eight weeks of debate.
Three: split day-one skills from trainable skills explicitly. Write both lists down. Use the day-one list to filter applicants. Use the trainable list as the 90-day ramp plan you give the new hire on their first Monday.
Four: record and review the loop. Use a tool like Metaview to capture interviews so calibration is not just a memory exercise. Onboard new interviewers by having them shadow real recordings, not hypothetical scenarios. Review the loop quarterly. Adjust the rubric.
What generalizes from Tally, and what does not
A few honest caveats. Tally is a small, healthy, growing company with engineering leadership that buys into hiring as a first-class activity. Not every company has that, and the moves Jan describes do not work if Eng leadership treats hiring as a tax. The first move in any team that does not look like Tally is getting leadership alignment that interviewing is part of the job. Without that, nothing else holds.
The second caveat: ranked rubrics and kickoff meetings are not a one-time setup. The rubric drifts. The panel turns over. New roles come up that look superficially like old ones but are not. The discipline Tally has is not the first kickoff, it is the hundredth. This is operating work, not project work.
What does generalize: the underlying principle that you can either design the hiring loop or inherit one. Every team is doing one of those two things. Most are inheriting. The ones who design tend to outhire the ones who don't, even when they are smaller and slower. If you take one thing from the Tally conversation, take that. The rest is implementation detail you can borrow from posts like this one or from 10x Recruiting, the show where we work through these problems with people who have actually solved them.
Bring Metaview into your hiring stack.
Live notes, structured scorecards, and ATS sync - set up in under 10 minutes.
Frequently asked questions
How do you actually get engineers to take interviewing seriously?
Make it part of the career ladder and budget for it in roadmap planning. If interviewing is a promotion criterion and Eng leadership accounts for the load when planning sprints, the question of whether engineers should care answers itself. Cultural framing matters too: senior engineers should be visibly engaged in hiring, and managers should refer to it the way they refer to code review.
What does a good interview kickoff actually look like?
Full panel in the room, list of competencies that could matter for the role, then a forced ranking down to five. Each interviewer leaves with a specific question area and a rubric for grading. The kickoff itself takes about an hour, and it removes most of the panel-debate-after-the-fact that drags loops out.
How do you decide what is day-one vs. trainable on the job?
Ask which skills are expensive to teach. Communication, empathy, problem framing, taste: those are expensive. Specific languages, frameworks, internal tooling: those are cheap. Hire for the expensive skills, teach the cheap ones in the first 90 days. Write both lists down so the panel knows what is filtering vs. what is ramping.
Why is "the process that hired you" usually wrong?
Because it was almost certainly not designed, it accumulated. A previous interviewer pieced it together for a different role, at a different stage, with different constraints. Without active redesign, every new hire reinforces a process that may not match the current company, and the bias built into it compounds over each cohort.
Where does AI fit into all of this?
AI compounds the work that teams like Tally already do. Structured rubrics, ranked competencies, and recorded interviews give an AI layer something useful to score against. Without that scaffolding, AI in hiring is a vibe-check accelerator. With it, AI runs sourcing against the agreed rubric, ranks applicants against day-one skills, captures interviews for calibration, and reports back on which signals predict hire quality.
