Bring shadow AI into the light. Calibre helps teams review employee-reported AI use, evidence and open questions. People own the decisions. Check by Andes.
Check by Andes helps organisations review AI use employees report against their own rulebooks and record human decisions. This demonstration uses synthetic data and the fictional company Larkspur Freight. Watch: https://t.co/4iThcVW1R1
Brett Schklar of The AI Adoption Institute on AI at work: "Change the culture, make it an open and free place for people to talk about what they're doing."
And to CEOs: "It's on you."
The Human Check, episode 2: https://t.co/d6mjdCzekK
Follow every AI use around one loop. When Legal asks for the same tool, the earlier decision is reused. Good use moves faster.
Every number on a board links to the cases behind it.
Open the demo (Demo · synthetic data): https://t.co/AvhoMPuFfR
Bring shadow AI into the light.
A 120 person food manufacturer found 384 AI apps in use. About 20 were sanctioned.
Brett Schklar: "Let's recognize that that is an opportunity and not a challenge."
The Human Check, episode 1, with The AI Adoption Institute. Write up: https://t.co/43bxq6KdSn
Someone tells you they've been using an AI tool for part of their job. Nobody in the room can answer the follow up questions yet.
Unknown is not no. It isn't yes either. It is a list of things to find out.
New piece: https://t.co/pCeHhfWrIj
Andes is the company. Calibre is its software for reviewing employee-reported AI use. Open the demo: https://t.co/AvhoMPuFfR The film: https://t.co/6ZYHjMtHqj Bring shadow AI into the light.
Every organisation asks two questions about AI. What is it worth to us? And what does it expose us to?
Most answer them on two separate tracks, in two teams, on two clocks. That split is the problem. A thread on running both in one loop.
https://t.co/QnOGaHx9hm
Asking again closes both books. Still using it? Did the benefit hold? One question tells you whether value showed up and whether the risk changed. Every number on a board links to the cases behind it. Human judgement stays visible.
Every organisation asks two questions about AI. What is it worth to us? And what does it expose us to?
Value and risk are two wheels in one system. Governance that keeps good use moving.
Calibre, by Andes.
https://t.co/hoFmkxjU4J
https://t.co/6ZYHjMuffR
FICTIONAL EXAMPLE: someone reports using an unapproved AI tool on public help pages. The useful outcome is not a reprimand. It is a clearer guide, a named owner and a reply to them.
The last time someone reported AI use at work, what changed?
https://t.co/jTMXLhc7gX
@callowayesq@vsachar "Verify" with no name attached tends to mean nobody. Once someone is assigned, what should they leave behind: a line saying the citations were checked, or a record of which authorities were opened, by whom and when?
@basilpuglisi Placing yourself by what you could show an auditor today is a useful test. Going from Tier 4 to Tier 5, what is the smallest record that counts: the named approver and the date, or also what that person reviewed before deciding?
@CKPillaiAI The third part is the one that gets tested later. When an approval is questioned months on, what should the record hold besides the approver's name: the evidence they actually saw at that point, or only the output they signed?
@vsachar The Butler Snow detail stays with me: a policy, a committee, and nothing between the draft and the filing. When a firm does add that step, who should sign the record of what was checked: the person who drafted it, or the lawyer whose name is on the filing?
@christinayiotis Useful list of activities. The harder part is often the record behind them: who signed off on a given use, on what basis, and when. Does the recommended practice say what that record should contain, or leave it to each organisation?
@NotTimKowal Good episode. One thing the bill text leaves open: it asks for disclosure "for all documents submitted to the court." Do you read that as only the filings where AI helped, or every filing once a lawyer uses it at all?