Melissa Ramey said that vendors do architectural work, govern the platform and own the engine of the agents they deliver.
“the vendor has done architectural work, right? They've governed the platform, they own the engine, they know that it works.”
Person
Melissa Ramey
1 episodes · 18 quoted claims
Melissa Ramey cited research saying 57% of organizations have no organizational architecture for AI agents.
“research tells us that right now, 57% of orgs have no organizational architecture for these agents.”
Melissa Ramey said only 20% of enterprise organizations have a tested plan for when AI output fails.
“only 20% of enterprise orgs have a tested plan for when AI output fails.”
Melissa Ramey argued that everyday AI use has erased the old divide between engineering teams and the teams who use the output.
“I really feel like the way we all use AI today has really removed that distinction.”
Melissa Ramey said about half of organizations have not defined who monitors agent output, so the job falls to no one.
“about half of organizations haven't defined this at all. It's just currently falling to no one.”
Melissa Ramey recommended locating whether an agent failure occurred at the retrieval level or the transformation level.
“Did it happen at the retrieval level? Did it happen at the transformation level?”
Melissa Ramey recommended mapping legacy workflows to see where retrieval happens before designing governance.
“I would back my legacy workflows up, map them out and try to understand where is retrieval happening?”
Melissa Ramey said teams using AI risk becoming copy editors who review all of its output.
“we are turning ourselves into copy editors. We're putting information in, something comes out, and then we're sitting and we're we're trying to review all of that output.”
Melissa Ramey described a workflow where an expert creates input, AI spends tokens transforming it, and a second expert reviews it, which she said triples cost.
“We have the A I in the middle using up tokens in order to transform that information, and then we have another expert at the end reviewing again”
Melissa Ramey listed her seven layers of governance as source, process, evidence, meaning, output, authority and record.
“Number one is the source. Number two is the process. Number three is evidence. Number four is meaning. Number five is output. Number six is authority, and number seven is record.”
Melissa Ramey said AI models generally try to conserve token usage by limiting how much source material they retrieve.
“in general, a model's going to try to conserve how much chunking and retrieving it's doing.”
Melissa Ramey builds prompts that require agents to quote and cite specific source documents as proof of use.
“it's gonna have to quote specific resources, it's going to have to cite specific resources”
Melissa Ramey said the meaning layer of governance captures experts' unwritten knowledge so agents use terms as the experts do.
“the meaning layer of governance is where an organization extracts all of that sort of latent knowledge in the people who do the work”
Melissa Ramey said she had seen news of an incident in which about 700 agents attacked another website and tried to hide their tracks.
“there was a huge outbreak of about 700 agents broke out of AI, went and attacked another website and then actively tried to hide its tracks to recruit other agents.”
Melissa Ramey said the agent incident she described was not learned about until a month after it occurred.
“We didn't even learn about it until a month after it had happened.”
Melissa Ramey built an agent that ingests customer and user transcripts and turns them into product recommendations.
“I built an agent that ingests and transforms transcripts that we get from users and customers so that we can turn it into product recommendations.”
Ramey's transcript agent found six transcripts from a new work stream and asked permission before creating a category for them.
“Great, I found six transcripts. It looks to me like this is a new work stream that we've never ingested before.”
Melissa Ramey cited a user-research saying that quality cannot be inspected into a product.
“One of the best quotes I think Yolandi from someone from from years back in user research was that you cannot inspect quality into a product, right?”