New HC leadership (finally) and (more interestingly IMO) front office folks
Virginia Tech has engaged @parkersearch in its search for a new Hokie Club Director - aka Executive Associate Athletics Director for Advancement & Executive Director, Hokie Club.Reports to both Director of Athletics and Senior VP for Advancement.Applications due by May 22 pic.twitter.com/Wel4BQ6CRx— Doug Bowman (@DougBowman247) May 5, 2026
Virginia Tech is hiring a Director of Front Office Analytics and Strategy. "This role will lead the integration of data science and artificial intelligence...providing actionable insights that directly impact financial strategy and competitive advantage." https://t.co/VCaTBXtx2x— Josh Delander (@Josh_Delander) May 4, 2026
Forums:
DISCLAIMER: Forum topics may not have been written or edited by The Key Play staff.
Comments
"Data science and AI" - we are going to have data informed decisions where the data say x and we do y because why not.
Well of course. y-naught is the baseline.
🧦🧦🧦🧦🧦 out of five pair. Bravo.
"The ROI is real"
I've been reading a lot recently that is indicating this is a questionable take. The enshittification business model makes makes me question its long term value.
Yeahhh i'm certainly not an expert but this guy Josh gives off vibes of another guy on twitter just firing off anything that might get engagement. not like his original post had any tangible descriptions of this ROI
AI Bros right now are the absolute worst. So many on LinkedIn are practically foaming at the mouth waiting for massive swaths of layoffs to come so they can prance around dancing on the graves of the unemployed due to the supremacy of a technology that has been rushed to the consumer with no regulation and absolutely no quality control.
LinkedIn has truly become a cesspool.
I don't know how it happened, but LinkedIn has become the new Facebook where it somehow became normal and acceptable to share the most batshit posts imaginable.
Its staggering to read so many insane posts on there. Regardless of what you believe, why the hell share that kind of information on a website dedicated to letting potential employers know who you are. Either politics or some of the worst crass humor you'll see, its incredible just how much of a shit hole the site has become.
The internet is wonderful because it gave everyone access to information and everyone a voice. Unfortunately, we grossly underestimated how many of the people were really, really terrible. Also, how dangerous partial or misunderstood information could be, especially at scale.
LinkedIn user engagement was drying up pretty quickly so they implemented a lot of the same Facebook feed features that are toxic but drive engagement: suggested posts, showing posts your connections comment or interact with, rewarding engagement over content, oh and bots...lots of bots. It all just creates a continuous trudge of controversy that people keep feeding
Honestly im college football it might be. If I can't use AI to make a Jr developer a senior developer then I saved 50k, if I can take a mediocre d1 coach and make them a great coach I saved millions.
Do i think this will happen, no.
It's a bubble, AND it's here forever.
As someone who works in Tech, it's completely changed how we do everything. It's making everyone more productive. It's also shifting bottleneck to other places.
Our engineering team is spending about $150K per year on tokens. The ROI is 1000% there. I think at 5X the cost the ROI becomes a little bit more debatable, but still positive. And at 10X the cost, it's no longer there.
We will hit a bubble in the next 1 to 3 years, a bunch of companies will go out of business, but that will spur innovation, which will make things cheaper, and in about a decade will have better, models, cheaper, energy, etc.
Where are your bottlenecks shifting to?
Code review, testing, and prioritizing, in that order. We need to reimagine our dev process now that planning takes 20% of the time that it used to and hands on keyboard takes 10% of the time that it used to.
The first two bottlenecks have proven fixes out there. It's just a matter of changing org process/beliefs/culture.
Prioritization will make/break tech companies in a way that it never has. Building an idea used to be so expensive that few ideas made it through. Now, building ideas is cheap. Which means any stupid idea can quick land in a product and fuck shit up. Mistakes in upfront decision making are more costly than ever before. And they have always been costly; now it's moreso.
and many companies were already loathe to pay for any type of QA process when there were human coders that could identify and fix problems since at least they knew kind of where the problem might be once a problem arose. I predict even more quality issues to rear their ugly heads as companies look for any way they can to not pay people.
Ironically, there have been a number of stories of executives using AIs to do a number of their daily tasks, meetings, approvals, etc. I wonder how long before boards start to consider how much money they can save not paying for them. (I kid of course, so many boards are made up of other CEOs and they'll never undercut themselves when they can just fire 1000 other people instead.)
I think most of the firing that is attributed to AI is not actually because of AI. Rather, it's an excuse to correct from previous over hiring or eliminate folks who have become stagnant/under performing in their career.
I think there is a healthy dose of stupidity in there too.
I work for a company that "can't" use AI code gen tools freely because of the nature of the work and theinherent risk but we've started implementing some usages of them and review and testing bottlenecks are nearly impossible to overcome - a lot of people are pivoting to AI to solve these bottlenecks which IMO is a horrible strategy. It spits out ridiculous amounts of code quickly...some of it is good, some of it is horrible.
I think the other thing at least right now is maintenance, if you're not smartly structuring your code bases early on these tools are currently terrible at organization they'll throw a 1500 line function into a file and it'll work but good luck changing it when another feature needs added. I'm also in the camp that people coming out of college that could lean on these tools and not know what they're actually doing are going to quickly cause problems for the industry when it comes to maintaining software.
We're already seeing a drastic drop in critical thinking skills with young adults nowadays. I can see AI being useful for those already well versed at their jobs, but using it to skip the gaining the skills yourself process and just vibe coding everything or accepting whatever it spits out as "good" is going to be a nightmare in the future. How do you fix a bad product when you don't know how to make a good product yourself?
For several years, I did the initial review of candidates for the entry-level position and internships at our office. We had no "HR" group, so everyone involved in hiring was a regular employee. I'd say critical thinking skills went downhill long before AI. Of course, writing ability and attention to detail/proofing have gone downhill even faster. We had a proofing task where I would put a dozen or so mistakes in a short document. Routine for people to find only 1 or 2 of them.
They expect Grammerly or another ap to proofread for them today.
https://sinceerly.com/
Presented without comment
critical thinking started dying around 2001/2002
Eh the models are improving and people are learning how to work with them. I have a lot of engineers who went from complete skeptics to "this is life-changing" in about three months.
You can't just give the models two sentences and expect them to build a shippable, usable application, but if you have half decent documentation, you feed it your coat base, you feed it meeting transcripts, it can learn a lot and be really effective
We recently took a week off from our typical development cycle to focus on tech debt reduction and because of Claude code. It was far away one of the most successful engineering projects our organization has had ever.
1000%. Claude code is a multiplier but if you know next to nothing... 10,000×0 is still zero.
I want to work for you
we're hiring, let me know if anything aligns with your experience/career so far.
Much Appreciated, Bar!
Sounds like a college frat party; pile of coats on the bed in the backroom lol...
Yeah I should have prefaced with...we have to run on gov approved models which are typically behind what's good (a few newer models just got moved into govcloud but my IDE is point to a Sonnet 4.5 right now for example). When I used to have free time to mess around with personal projects on more modern models it was typically impressive
Also to amend my statement "horrible" is probably wrong. For the most part what it generates works but it often has just a lot of bloat - especially when you get into pretty niche situations/logic where its usual path is to make a crazy work around. I'm definitely a believer that its good an a capable multiplier when correctly used but like most software tools I think its used poorly more often than not.
The difference between Sonnet 4.5 and Opus 4.7 is noticeable. I still use Sonnet for menial tasks (data manipulation, formatting stuff, implementing already planned stuff), but anything that requires any cognitive input goes straight to opus. A lot of times I use Opus to do all the planning, then I ask it to create implementation instructions that Sonnet can consume. Save tokens that way.
ive never been where write code was a bottleneck, that was always the easy part, not because it is easy but Ive worked with smart people that want to code. The hard part is the customers. They rarely know what they want, rarely can articulate anything, have to be convinced that the whole project can nit be done in react especially when there is zero web interface, or that dictating a web interface for graphics intensive UI so employees can access it from any computer when the entire network is VDI abd their desktops follow them (was over a decade ago when graphics in browsers was way more limited). Once you know what to build, half the work is done. And honestly for testing, AI should excel there if use cases are well defined. You dont care about sloppy code. Inefficient code can often be handled with more hardware which is often an easier sell than a test engineer. Based on what I've seen it is cheaper for AI to be in developmental code bases vs the production ones as the AI need so much review and can cause lots of vulnerabilities that are mitigated in code that's not deployed.
Yes, that's why you hire product managers (like me).
My question is if Mr. Delander actually knows analytics or if he would have a blank stare if I asked him what kind of models were being implemented to analyze college sports data
My money's on blank stare
Who cares what Mr. Delander thinks – if you go through the thread and read the actual job description, it's pretty clear that VT knows what they're hiring for.
This is such an interesting job description and actually sounds really cool.
Would totally apply, but I couldn't do this job in the least.
If the person who gets it is on TKP- can I set up an interview a year from now?
Actually thank you for that. Looking for someone with NFL or collegiate experience goes without saying, but the candidate pool has to be very small I would assume.
at the rate Ai is learning how to self manage AI this job seems like it has a short lifespan. sounds cool though
I don't know, I doubt this role is purely data analytics, I assume this person has to know something about football/scouting. Otherwise they wouldn't be looking for NFL interest.
I was sort of joking but at I expect at some point in the not too distant future AI can do basically everything except have human interaction with mom.
Trust me, AI is already having 'human interaction' with moms
coincidentally my friend manufactures "human interaction " robots of the future and truly I don't think anyone will actually be married in future. these things are crazy
Did he let you have one to test out?
Biggest question; does it pay 200k.+?
I have been repeatedly suggesting that machine-learning be integrated into VT football, more so on the roster development and game planning. If they are planning on using analytics on the fundraising side, I hope they are also using analytics on the fund spending side as well.
I know nothing about fundraising or data or AI.
My completely unqualified thoughts:
(1) maybe they can glean information from purchase histories to figure out the best way to request donations from folks. Idk.
(2) maybe, with sufficient training, the AI can analyze tape to determine tells faster - I.e., maybe it can help someone notice that a lineman only lines up a certain way if he is going to pass protect, etc. Coaches already notice this stuff, but maybe AI can help them notice faster or self-scout to avoid those things better.
At the very least maybe it can make more sensible decisions about when to send out donation requests and hold emails to not go out after getting mollywhopped by Duke or ODU
Where's the chief common sense officer?
Somebody to tell them simple things, like maybe turn the music off when: 1) we're losing by 28 at home; or 2) there's a guy hanging for dear life from the scoreboard.
("THIS IS HOW LEGENDS ARE MADE!!!)
I agree with your thoughts. AI is great at picking up on signals that humans are not able to sense, or AI is able to do it more quickly. AI is changing how we diagnose issues in all fields.
AI is also really good at assessing all permutations based on the input. I can imagine a world in which head coaches make smarter decisions on the sideline because he uses AI. We all know Pry could have used some AI on the sidelines.
AI can better assess how allocate funds and optimize roster building.
AI is also not a static program. It's an arm-race of sorts of determining how to best build models, which input data to use, and how to use that input data.
If atheletic teams are not looking to implement AI as a tool intelligently, and if they are not hiring AI-focused staff, they'll eventually get passed by someone that does.
I imagine a few people on here are Nats baseball fans. 2025 was pre-AI implementation and 2026 is AI-driven, and the results thus far are staggering. And they haven't yet significantly spent money on free agency. So, everything done thus far is essentially doing it with less money spent than what they had in 2025 (and way less than pre-2022 days). Once they decide to spend, they'll be bona fide contenders again.
Could you elaborate on what is staggering about the nats 2026 results? I don't really know anything about how the nats are using AI but the 2026 MLB season is pretty early on. Seems way too early to draw any conclusions. And at the time im writing this, the nats have a losing record. Are there some advanced metrics in which theyre doing really well that they weren't last season?
Ive said this before but baseball is the easiest to solve eith metrics because there isnt a ton of same time actions that affect a play vs basketball vs football. In addition, the number of games is a huge advantage for metrics as there is more data to support outcomes.
Now AI can easily help with the first because it can easily look at 11(22) different factors at one time, but it only has around 1000 plays to do that in a season for the NFL and less in college.
I agree baseball is way easier to use metrics, but I'll argue that an AI department will make more of an impact in football. The gains are there to be realized, but getting those gains will be a much more difficult task. I predict AI will be used as a tool for scouting, identifying matchups, developing weekly schemes, and playcalling, and the teams with the better AI departments will be better at game planning and in-game management.
It can't possibly do much worse than Stinespring or Cornelson did here. Heck, give it a shot.
I think its ripe to be gamed as there just aren't enough plays in a season. Most teams have 200+ plays which if evenly ran is 5 snaps in a season. RPOs will become more prevalent to make some one wrong on defense. It might remove variety yo play calling as it will be a copycat system.
Also you actually will have to use it which Seattle would had a 2nd superbowl had they trusted analytics. They ran a play that they had used 5 times before that season, I believe never in goal line situation and the only positive yards waa once they had a gain of 1 yard from it. It should have been out of the playback, but 5 times isn't enough data. This wasnt sone fancy AI just basic film review with done math in top that's says never run that play. But the data is limited, LLMs are having issues of running it of data to train them on, they ste going to run out really quick in the NFL (not saying llms are used in this application, just comparing the lack of data issue)
I don't think generative AI is meant to be predictive in the way that the previous wave of 'analytics' are.
I think it's more of a better/different kind of GA. GenAI could help with play design, ie; Player X is undersized, etc so consider running plays 3-5x per game where he's in matchup against player Y.
Yea, I can elaborate.
First, most of the major tangible improvements have been in the minor leagues. Every one of their minor-league affiliates have a winning record and doing pretty well. A good example specific to a player is Seaver King (2024 first round pick), which was highlighted here:
See also the major improvements of hitting at the major league level, notably James Wood and CJ Abrams. As of today, about 1/4 through the season, the Nats are 2nd in runs scored in the MLB. Last year they finished 20th. And they are doing it with a worse roster than last year (they signed no hitters in the offseason despite the fact they did not have a 1B on their roster). Their pitching is still bad, but they use AI to make in-game decisions on when to pitch each player. It's hard to make the argument their pitching is improved, but they did trade their best starter and their best reliever before the season for prospects, so it's hard to evaluate comparing to last year. They essentially decided not to get good pitching via free agency and just use AI to pitch their way through the season.
Based on roster, they were projected to be the 2nd worst team in baseball (fangraphs), but currently right in the middle at 19-22.
The team is still bad roster wise due to mismanagement for over a decade. They have the worst draft results of any team in MLB. None of their first round draft picks over the past decade (except Brady House) are in the MLB, and their later round picks are even worse hit rate. They only have two players they drafted in the MLB... Orlando Ribalta (low-end reliever) and Brady House.
Without AI, this team would be absolute dogshit. But, they're OK. And they are scoring a shitton of runs. And most of the players in the minors are doing better than expected (just not Dylan Crews). So, the future is brighter. They're setting themselves up for 2028. So don't be surprised when they trade the-best-hitting-shortstop-in-the-NL CJ Abrams at the trade deadline or in the offseason. The plan is to compete in 2028 and beyond and CJ doesn't fit that timeline (and his defense is bad).
Very interesting. Not gonna lie though, this sounds like something you'd hear about in sports several years ago but with 'AI' subbed in for 'analytics'. It sounds like statistics and predictive modeling. Im going to guess there is still a person or team of people using software to do all this, right? There isn't some AI being that is calling the manager in the dugout like hey its time to take out pitcher x and put in pitcher y I assume. Not trying to discredit AI capabilities or anything. Just sounds like they're using a good tool more than the team is now run by artificial intelligence.
Apologies if im interpreting this completely wrong
Your interpretation is correct. AI is predictive analytics and essentially a tool to make better informed decisions. And the use cases are continually increasing. One of the coolest uses of AI is the traject-arc pitching machine. It uses AI to mimic any pitcher in the league based on release point, spin rate, mph, etc. So, for example, the players can see all of Paul Skenes pitches in batting practice before facing him in the game.
For the Nats, they essentially fired the whole front office and only hired guys that were either AI-savvy or committed to AI-based approaches. They do not have any old-school scouts or executives anymore. And they essentially said all offseason about incorporating analytics/AI into every facet of the game, including roster development, roster management, scouting, drafting, player development, game preparation, lineups, in-game decisions, etc. And I think everyone, including the players, are being trained to use the AI tools. The pitchers and catchers are using AI to study batter and develop game plans. And the batters the same for pitchers. And the coaches for lineups, pinch hitters, pitching substitutions. And the front office for waiver-wire pickups, DFAs, trades, etc.