The back field where a career becomes a file
MLB teams use AI for scouting in 2026 on the same back fields where scouts have always lived. Dust skates across the warning track. A bucket tips over. A teenager spins a ball in his fingers like he is hiding something.
Bill watches the shoulders, not the radar gun. A faded Dodgers cap sits low on his forehead. Ink stains his thumb the way it has for forty years. A younger analyst stands beside him with a tablet held like a shield.
Two iPhones sit on tripods behind the mound, angled like a cheap broadcast. Another phone sits near the plate. The screen runs, and the pitcher turns into a moving skeleton. Dots mark hips, elbows, knees. Lines connect everything, joint by joint, like the body became a wiring diagram.
Bill keeps his eyes on the kid instead. He counts the breath between pitches, watches the face after a miss. He still wants the handshake when the session ends.
The front office wants more now. It wants warning signs, probabilities, and an explanation it can defend when a pick becomes a million dollar bruise.
So the tension sits right there in the dirt. Does the tech sharpen the scout’s instincts. Or does it tempt the room into thinking it already knows the player.
The new rule that made every pitch a decision
Spring 2026 brought a fresh kind of noise. The automated ball strike challenge system arrives for the regular season, and teams treat it like a language they have to speak under stress. MLB’s published rules are simple on paper: two challenges per team in regulation, with an extra innings safety net so a club is not left helpless.
That simplicity is a trap. A catcher now manages a resource. A pitcher now reveals temperament with one tap. A hitter now decides whether to fight a borderline strike or save his last challenge for a moment that can flip a game.
Clubs drill it in camp like it is a rundown play. Coaches stand behind cages and talk through leverage. Analysts chart who challenges out of anger and who waits for the pitch that matters.
Broadcasters had to adjust too. League guidance pushed networks to remove some strike zone graphics so players could not get a free hint on challenges.
This is what 2026 feels like. Baseball did not get less human. It got more exposed.
Three fronts that define modern evaluation
No organization wins by chasing one gadget. Smart teams build a stack and keep it grounded in baseball reality.
One front lives in the body. Computer vision captures biomechanics anywhere and flags patterns tied to injury and repeatability.
Another front lives in the game. Optical tracking and radar turn defense, swing decisions, and pitch shapes into searchable evidence.
The last front lives in the brain. Language models let staff talk to their own archives and pull old truths back into the room without digging through ten thousand PDFs.
A quiet filter decides which tools survive. Teams ask three questions before they trust a new layer. Can it travel or predict something that matters. Can it help a person make a decision faster without making that person lazy.
With those questions in mind, here are the ten building blocks shaping scouting right now.
The 10 building blocks of AI scouting in 2026
10. Two phones turned biomechanics into a travel kit
Biomechanics used to mean a lab, reflective markers, and a budget that made small market clubs flinch. Uplift Capture broke that barrier with a blunt idea: two iPhones, two tripods, and computer vision good enough to build a 3D model.
The overlay is what startles people the first time. A pitcher becomes a stick figure with tracked joints and angles. The animation looks clinical, almost cold. Coaches love it for one reason. It gives them language when the body lies.
Bill still writes “late arm” sometimes, because that is how scouts speak. The younger analyst points to sequencing, to trunk timing, to the instant the chain breaks.
Then the conversation turns darker. Forearm Flyout is the phrase that makes scouts lower their voice. Bill explains it like a warning you give a friend, not a statistic you sell to a room. When the forearm flies out early, the elbow gets dragged into places it cannot live for long. The kid might survive it for a year. The kid might survive it for a month. The red flag is not a diagnosis. It is a reminder that velocity can come with teeth.
Portability is the real story. Dominican academies can run the same setup. Remote complexes can use it too. The tool does not replace the scout there. It follows the scout into places that never had a lab.
9. The challenge system changed what teams want from catchers
Framing used to be currency. The challenge system changed the exchange rate.
A catcher can still steal strikes in the moment. He can also give them back if he burns a challenge on a pitch that never mattered. Pitchers show their cards too. Some chase fairness. Others stay calm and save bullets for leverage.
This is why scouting meetings now include a new phrase: challenge discipline. Clubs grade communication and composure like tools. Calm is a skill. Timing is a skill. Leadership shows up when the dugout wants to tap the helmet out of frustration.
Bill watches those moments the way he watches swings. A glove can be trained. A steady heartbeat is harder.
8. Trajekt Arc made timing a measurable skill again
Hitters used to rely on memory, a scouting report, and a prayer. Trajekt Arc changed the cage into something closer to a truth serum.
The machine projects a life size image of a pitcher and delivers a real ball from a realistic release point. Players do not just “see velo.” They see the delivery, the slot. They see the deception.
Teams pay real money for that realism, and not because it looks cool on social clips. Coaches use it to test whether a prospect’s timing survives when the visuals get honest. One session can answer a cruel question fast: does he load too late when the ball looks big league.
A good staff treats the reaction as data too. Some hitters stiffen up when the video looks too and start chasing early. Some fight through it and adjust.
That adjustment matters more than the first swing.
7. Hawk Eye and Statcast turned every rep into evidence
A scout used to win arguments with memory. Tracking replaced that advantage with receipts.
Hawk Eye style optical tracking now follows player movement and ball flight with enough detail to change how teams talk about defense, baserunning, and even body positions at release. Every park becomes a camera grid. Every game becomes a file.
The practical impact shows up in small moments. A shortstop takes one false step. The system marks it. A coach clips it. A development plan changes.
Front offices no longer argue about what happened. They argue about why it happened, and they can rewind the exact frame where the answer starts.
6. Defense models changed what scouts even notice
Defense used to live in adjectives. Smooth. Twitchy. Instinctive. Tracking made those words less powerful. AI made them less necessary.
Models estimate route efficiency and reaction time. They surface plays the eye forgets, the ones that look routine until you compare them to elite baselines. A center fielder can take a perfect angle and still lose a ball to wind. The model credits the route. The staff still asks whether the first move was late.
Numbers can also create a trap. A team can start worshiping route graphs and forget the chaos of a real game. Great organizations keep film in the middle. The model points to the play. Humans decide what the play means.
Bill likes that order. It keeps the room honest.
5. Pitch design tools started scouting a future pitch, not a present one
Velocity still sells. Shape sells too now.
Teams model pitch movement and compare it to buckets that succeed in the majors. A pitcher with one great fastball can become a different prospect if the staff believes it can teach a slider that pairs with the fastball’s ride.
This is not magic. It is development planning dressed up as scouting.
The bet changes because the questions change. Coaches look for hands that can learn a grip. Analysts look for release traits that support a new shape. Scouts look for a kid who will do the work when the results do not show up immediately.
Bill cares about that last part more than any graph.
4. Injury risk models became draft room tension you can feel
Every team has a scar from an arm that broke. Draft rooms learned to fear the medical report as much as the radar gun.
Risk models did not erase that fear. They gave it shape. A score pops up. A red flag appears. The room gets quiet.
Good teams avoid lazy fear. They do not let a model erase a player. The warning becomes a prompt that forces better questions. They order extra looks, talk to trainers. Study recovery habits. They ask about sleep, routine, and how the player responds when his body feels off.
A scout’s job changes here. He must describe not only the arm, but the choices the player makes when nobody is watching.
3. In house language models turned the archive into a teammate
The modern problem is not information. The modern problem is finding the right piece of it before the meeting ends.
Front offices are building in house LLM layers that sit on top of their scouting reports, player development notes, medical logs, and tagged video. Some clubs give the tool a nickname that sounds like it came from a clubhouse joke, something like Scout Chat or Report Room. The name is not the point. The workflow is.
The prompt sounds like scouting again. “Find me a low spin lefty with a heartbeat like Logan Webb.” “Show me hitters who tightened chase after we changed their load.” “Pull every report where we worried about late life on the heater.”
Under the hood, the tech is rarely glamorous. A team with a Databricks lakehouse can wire retrieval straight into its internal data platform and let non technical staff ask questions without writing SQL, the same direction that clubs like the Texas Rangers have discussed when talking about modernizing their analytics stack and making data usable across the building.
Vendor stacks show up too. Some teams prototype the search layer with tools like Vertex AI Search and Gemini, or Azure OpenAI style services, then wrap it in strict permissions so a scout only sees what he is allowed to see.
The value is not that the model decides. The value is that the room stops forgetting. The tool pulls an old report into a new debate and saves a staff from repeating its own mistakes.
2. Comparison models risk standardizing the perfect swing
Comps used to be storytelling. Data sharpened them. AI can turn them into a trap.
When every club leans on the same Statcast baselines and the same public leaderboards, the sport starts chasing one approved swing. The “perfect” shape gets rewarded. The weird winners get questioned.
Models can widen the map if teams use them with imagination. An outlier swing can look ugly and still produce hard contact in the right zones. A pitcher can move like a puzzle and still create late break that hitters hate.
Smart groups build friction on purpose. One voice argues for the model. Another voice argues against it. Bill usually gets the second job, and he does not soften his words when the room starts falling in love with a clean comp.
1. The best organizations still scout the person, and the tech makes that harder
No model can sit with a player after failure and measure what comes next.
Bill watches the handshake because it still tells him something. He listens for how a kid answers a tough question without spinning. He watches whether the player asks for extra work after a bad day or disappears.
Tech complicates belief. A green light can seduce a room into silence. A red flag can give a room permission to walk away.
That is why the best organizations keep humans in the loop and make them be specific. A scout cannot say “tough” anymore and expect applause. He has to describe the moment, name the behavior. He has to show how it showed up under stress.
This is where the technology finally breathes. It does not replace the scout’s voice. It forces the voice to get sharper.
What 2026 is really testing
The next edge will not come from another camera. Governance will decide who wins.
Teams will fight over data rights and privacy, especially when a player’s movement profile starts feeling like property. Bias will matter too. A model trained on players with access to high end tech can punish the ones who grew up without it, even if they can play.
Baseball will still reward the outlier, because baseball always has. Some swings will look wrong until the ball jumps, pitchers will move like a mess and still get outs. Some kids will fail the camera test and still win games.
A 3D motion capture demo makes that tension easy to picture. The skeleton overlay looks like truth. The real truth still lives in what the player does after the skeleton says something scary.
That leaves the question every scout feels, whether he admits it or not. As AI becomes a standard part of evaluation, will teams keep room for the weird winners. Or will the sport train itself to ignore the next one because he never fit the template.
Read More: MLB First Pitch Traditions: The Most Iconic Ceremonial Tosses of 2026
FAQs
Q1. What does AI actually change in MLB scouting in 2026?
A1. It turns video, tracking, and reports into fast answers. Teams still scout the person, but they arrive at the meeting with more proof.
Q2. How does the ABS challenge system affect scouting catchers?
A2. Teams now grade challenge discipline and calm. Catchers can lose value if they waste challenges in low leverage moments.
Q3. Are teams really using iPhones for biomechanics now?
A3. Yes. Portable setups let clubs capture 3D movement outside a lab, which matters in remote complexes and academies.
Q4. What is “Forearm Flyout” and why do teams care?
A4. It is a movement red flag tied to stress on the arm. Teams treat it as a warning, not a diagnosis.
Q5. Can AI models miss great players?
A5. Absolutely. Models can reward “clean” comps and punish outliers, so strong teams keep humans arguing with the machine.
I bounce between stadium seats and window seats, chasing games and new places. Sports fuel my heart, travel clears my head, and every trip ends with a story worth sharing.
