Valentina Akpan: Founder, Rellatech. Admin and technical virtual assistant for businesses and startups with teams.
The largest AI company in the world just shipped a product for putting agents to work, and in the same breath shipped a long list of reasons not to let those agents work on their own.
On July 22, OpenAI announced Presence, an enterprise platform for deploying AI voice and chat agents across customer support, outbound sales, and internal service requests. As Help Net Security reported, OpenAI describes Presence as a deployment platform rather than a standalone model, bundling company policies, guardrails, approved actions, simulations, and evaluation tools into a single system that a business can actually operate.
The details are worth reading closely. MLQ News reported that deployments are led by OpenAI's own Forward Deployed Engineers alongside a short list of global systems integrators, with BBVA Mexico and SoftBank Corp among the launch customers. OpenAI's Codex agent reviews production interactions and proposes improvements, but staff members are still the ones who have to test and approve every change before any of it goes live.
Presence did not arrive alone. A July 23 analysis from Being Shivam counted four major agent platform launches in roughly a month: OpenAI Presence, Meta's Business Agent Platform, the NVIDIA and ServiceNow Project Arc, and Google's Gemini Enterprise. The piece makes a sharp observation that is easy to miss under the marketing: every one of these platforms now leads with guardrails and audit controls, which means those control features have stopped being a selling point and have become the price of entry.
The same pattern shows up across how these products are sold. The companies with the most to gain from selling autonomous agents are the ones being most careful about never letting those agents run unsupervised. Escalation rules route conversations to humans, approval steps sit in front of every behavioral change, and engineers are embedded in every deployment. When OpenAI sells an agent into a bank, it sends people along with it because the software on its own is not the product.
That tells you where the real cost sits in a real deployment. The model is the cheap part, and the expensive part is the trust layer around it: the written policies, the review workflow, and the person who is genuinely accountable when the agent gets something wrong. Enterprises pay for that layer by hiring dedicated engineers and integration teams, but a business running with under 15 people does not get a Forward Deployed Engineer showing up to hold that layer together.
What a business of that size gets instead is a person who actually knows the operation, sets up the automation, watches what it produces, and catches the wrong booking or the half right reply before a client ever sees it. That is the shape of my executive assistant services and my CRM and automation work: real tools, including AI where it earns its place in the workflow, with a human running the system around it.
What to do with this news
The agent platforms will keep launching, and that is fine. Watch them, borrow the useful ideas, and notice that every single one is built around supervision. If your team is losing its week to inboxes, scheduling, CRM upkeep, and follow up, you do not need an enterprise deployment. My monthly retainers give businesses and startups with teams a dedicated assistant who runs the day to day, brings in automation where it fits, and keeps a person watching the work. If you want to see what that looks like for your setup, book a consultation and I will map it out with you.
Contact MeNot ready for a call? Take the free operations assessment and I will send you a written breakdown within 24 hours.
Related Services