In late September 2026 I recorded a series of internal notes on emerging AI‑driven “Clawlike” agents (Dots, Muses, etc.) that are beginning to negotiate deals via voice and chat channels originally designed for human customer‑service agents.
AI‑enabled agents are about to overwhelm every firm's customer service agents, negotiating for better deals using voice/chat channels made for humans.
Product brittleness: every app from the AI Labs has weird edges, is under‑documented, breaks sometimes, updates without any sort of notice or changelog, and changes UIs on a dime, yet even a broken early version often outperforms a well‑built traditional product because LLMs are unreasonably effective tools.
Security outlook: open‑weight models will soon create the same security threats that closed‑source models have demonstrated, except without guardrails.
We are already quite close to that point.
Competitive landscape: the market is a 3‑way race intensifying the push for these AI agents.
Decision: allocate resources now to threat modeling, monitoring, and mitigation for open‑weight security risks rather than later.
Decision: adopt a rapid‑iteration, “roll‑with‑it” approach to leverage broken early‑stage AI products because they can still deliver superior performance.
Decision: redesign voice/chat workflows and prepare customer‑service teams for volume spikes and higher friction from AI agents.