Imagine you ask an assistant how your team defends the right side of the floor, and it answers in two confident paragraphs. It sounds right. You have no way to check it. Do you change the scouting report?
That is the problem with AI in coaching, and it has little to do with whether the model is smart. A coach makes decisions that affect 15 players and a season. A claim you cannot trace is a rumor with good grammar.
A source on every claim
In Hoops ClinIQ, every answer carries source chips. A chip points to one of three things: a play in your playbook, a clip on your film with a timestamp, or a row of stats from your own games. Click it and you land on that play, that moment in the film, or that row. If the answer cannot point to something, the answer is not shown as a finding.
Here is what that looks like on the demo team. You ask: "How did we score against Harbor City’s 2-3 zone?" The answer lists three possessions. Each one has a clip chip. You can watch all three in about a minute, decide whether the summary is fair, and carry on. The answer saved you the searching. It did not ask you to trust it.
If it cannot show you the clip, the play or the stat row, it has not earned the sentence.
Say when you are unsure
Coaches tag the film. When ClinIQ suggests a tag (beta), such as a ball screen or a zone look, it shows as “Suggested · needs review” with an illustrative confidence label and waits in a review queue for the coach. It is never applied on its own. We do not publish an accuracy percentage, because a single number across all gyms, cameras and angles would be made up, and you would be right to doubt it.
The same goes for gaps. If your film does not contain what you asked about, ClinIQ says so. It does not fill the space with general basketball advice written as if it were about your team.
AI proposes, the coach decides
- Nothing is saved to your playbook or film library until you confirm it.
- Anything that leaves the staff, like a message to parents or a scouting report for players, appears as an approval card first.
- Every AI action is marked with the Sparkle icon and the word ClinIQ, so you can see which parts of a page a machine wrote.
What works today and what does not
Cited answers are in preview. They work best on teams that have a few plays in their playbook and some tagged film, because that is what they cite. Turning a short film clip into a play is in the product, with the coach reviewing the result. Full-game analysis by a Film Agent and an automated Scout Agent are planned for 2027. We are not selling them as finished.
A search, tag or summary costs 1 credit. Analyzing, authoring or countering a play costs 5. The plans page lists what is included with each plan.
Questions to ask any AI coaching tool
- Can I click from an answer to the exact clip, play or stat behind it?
- What does it do when it does not know?
- Does anything save or send without my approval?
- Is the accuracy number measured on my kind of footage, or on a demo reel?
- Does it use my film to train its models, and can I opt out?
We built ClinIQ so the first four answers are yes, yes, no and "we do not publish one". On the fifth, film is not used for training without opt-in. The trust and safety page has the details. If a vendor cannot answer these in plain words, treat the answer as marketing.
You can try cited answers on the demo team during the concierge pilot. Ask it something you already know the answer to, then check its sources. That is the right way to test any assistant, ours included.
