> ## Content Index > Fetch the complete content index at: https://intelligence.artificialrealitymedia.com/llms.txt > Use this file to discover other available public pages before exploring further. # Claude wrote the specification. A human still had to deploy it - URL: https://intelligence.artificialrealitymedia.com/specification-to-deployment-gap/ - Published: 2026-09-02T12:03:00.000Z - Updated: 2026-09-02T12:03:00.000Z - Description: A health practitioner built fifteen AI skills and could only fix the skills. The gap between a machine-written spec and a running system is the product. - Author: Galina Fendikevich - Tags: Health & wellness, #For AI Companies, #What needs to be built, #Import 2026-09-03 05:20 The product gap Dr. Nikki Siso is describing is the distance between a machine-readable specification and a running system. [Claude](https://claude.ai/?ref=intelligence.artificialrealitymedia.com) wrote her the step-by-step build instructions for her fifteen-skill health assessment. She handed them to a human developer, because nothing takes a specification and deploys it. Her question to Artificial Reality is the one builders should answer. > "Claude gave me the instructions. I handed the instructions to the developer. If I can get the step-by-step instructions from Claude, why can't I just hand this to a model and say: build it?" ## Where does the line between operator and engineer actually fall? Further along than the usual framing assumes. Siso, a certified holistic health practitioner with no technical training, drew it precisely: > "I was doing it on my end, because otherwise I was finding a bug, but my developer had to fix it. And now I could do it myself. But I couldn't fix what he set up. I couldn't fix if the conductor and the health assessor weren't talking together. That was on the back end. I could only fix the skills." Her verdict on that boundary: "But the back end could still be way easier." The content, the tone, the domain logic and the individual agents were hers. The orchestration between agents, the deployment, the token budgeting and the delivery mechanism were the developer's. The received wisdom is that domain experts produce content and need engineers for anything real. This operator produced the substance of the product and needed an engineer for the plumbing. ## Why is voice still not deployable for her? Latency, and she gives a number. She wants the assessment conducted in her own voice and will not ship it: > "I would love if AI could actually use my voice in real time to have a dialogue with someone. Because currently — yes, there's [ElevenLabs](https://elevenlabs.io/?ref=intelligence.artificialrealitymedia.com), there are ways to do voice cloning, but the speed of the conversation, there's a very long delay." > "That reply needs to be read, interpreted, and then response created. That happens pretty quickly on type. It does not happen fast enough on voice. Right now it's probably a good five to eight seconds." She describes the effect as "jerky," and says closing it would make the assessment feel "more connected." Five to eight seconds is a specific target, and it is the second time in Artificial Reality's Season 1 research that an operator has wanted a voice interaction and declined to pursue it on latency alone. ## What does she mean by tone that does not fade? She is asking for tone that persists across a long interaction rather than degrading over it. Her requirement is a system carrying "a really strong sense of my tone that sticks and stays, right? That doesn't fade." The word fade is doing the work. For a coach, how something is said is part of whether it works: > "The way I say things or the way you say things might land for someone differently. You could have heard the exact same thing from someone else and it didn't register. And then when I say it the way I say it, somehow you're like: oh my God. So capturing that tone is very important per coach." The cost of not having it is manual: even her automated text check-ins need her to "go through every single text message, make sure it's all there, it's all like me." That is hours of maintenance on something already described as automated. ## What should not be automated? Accountability, on her account, and this is a limit rather than a feature request. Galina Fendikevich raised the client side of it: automated check-ins are easy to ignore. Siso's diagnosis: > "There has to be a human at some point in the chain. So even if AI did the automatic prompting of, hey, what'd you eat today — your reply, if you knew, was gonna get seen by a human, then it challenges your integrity to not reply. That's what's missing." > "When it's just a hundred percent AI, then you're just like: I can ignore it today." What she does automate is the monitoring, not the relationship: when a client on her detox program logs a low score, the system alerts her. In her words, "I get the prompt when they need me." ## The four products this operator is describing - **Specification to deployment.** Take the build instructions a model already produces and stand the system up, without a person in the middle. - **Back-end orchestration a non-engineer can repair.** She could fix the skills, not the wiring between them. - **Voice with conversational latency.** Her stated blocker is a five-to-eight-second gap, not fidelity. - **Tone that survives the whole conversation.** Captured per practitioner, and audited without reading every message. ## What a founder should take from this Do not build the assessment. She built it, in fifteen skills, with no technical training, and she does not want your version of her domain expertise. Build the boring half she cannot reach: deployment, orchestration, delivery and token budgeting, sold to someone who has never opened a terminal. The specification for that already exists in machine-readable form, written by a model, sitting in a folder waiting for a human to type it in. That is the gap, and it is not a gap in intelligence. ## Where this comes from S1E7: How a non-tech founder built a 15-agent AI business on Claude — the full interview with Dr. Nikki Siso. Listen or watch: [YouTube](https://www.youtube.com/watch?v=R%5FzWAgAEIMw&ref=intelligence.artificialrealitymedia.com), [Spotify](https://open.spotify.com/episode/08s9u7LSBTtGGx94b4z2MZ?ref=intelligence.artificialrealitymedia.com) or [Apple Podcasts](https://podcasts.apple.com/us/podcast/how-a-non-tech-founder-built-a-15-agent-ai/id6780699025?i=1000777283240&ref=intelligence.artificialrealitymedia.com).