Start with the work that keeps piling up
Buyer follow-ups, order preparation, cost checks and project coordination consume time before anyone makes the important decision. One agent with a defined job can prepare that work from your existing records.
The goal is more useful capacity from the team you have. It is not a promise that every business will avoid hiring or save a fixed number of hours. Pick a job with enough repetition to learn from and a result you can check.
Teach the reasons behind your decisions
Uploading the price sheet is a start. Explain why one customer gets different terms, why a cheaper supplier is sometimes the wrong choice and which exceptions always need you. Those explanations make the context useful.
For each agent, name the job, the manager, the records it may use and the decisions it must bring back. Review and correct the first outputs. You are teaching how the business works, not just giving it files.
The first offer: buyer follow-up
The first offer is one organization, one Sales and Key Accounts agent, one defined queue, one email connection and one governed reference pack of approved account, product, pricing and commercial-rule information, with a named process for keeping it current.
For each buyer message in the queue, the agent brings together the relevant history, identifies outstanding commitments, prepares a sourced account brief and a reply draft, and flags missing information or terms outside your approved rules. A person on your team reviews and sends every reply. The agent does not send on its own, run outreach campaigns, make binding commitments, place orders or post to your books.
How a pilot runs
Pilots are paid, and access is by request during private preview. We agree the price, the measures and the decision at the end before any work starts. Your pilot runs in a dedicated deployment that thinqOS hosts and operates, and live data arrives only after its isolation has been verified. We measure your current process first, then run the agent for 60 days and review the results with you. After the read-out it continues month to month unless you cancel.
The pilot is on the Small Business plan at $250 a month (USD, taxes extra) with a $2,500 one-time launch for the one-role setup. It includes $75 of managed AI a month, valued at the customer usage price of provider cost plus 30%, and a spend cap of $225 a month on total managed AI usage, including that credit. We warn you at 80% of the cap, and usage stops at the cap until you raise it. We make no promise about results; measuring them is what the pilot is for.
A morning with buyer follow-up
Illustrative workflow: a retailer sends an order earlier than last season. The Sales and Key Accounts agent checks the account history and your approved reference pack, notes an open commitment from the last call, and prepares a brief and a reply draft. Stock or cost data it cannot confirm is flagged, not guessed.
Your team reviews the draft and sends it. A correction is kept as context to check on the next comparable order. This is a proposed workflow, not a measured customer result.
One team, clear responsibilities
Each person has their own Mind. Specialist agents can handle sales, operations or another defined job. The company Mind holds shared business context for authorized use. A specialist is worthwhile when its work has distinct goals and recurring decisions; you do not need an agent for every tool.
Choose who maintains the agents after launch. Your own team may do it, or you may want help with setup and ongoing improvement. Agree the responsibilities before the work starts.
Measure before expanding
Begin with one workflow and a small set of representative records. Check factual errors, preparation time, manager review time and whether work actually gets completed. Time released is different from cash saved.
Expand responsibilities when the evidence supports it. Keep a person able to continue the work if an agent or connection is unavailable. A useful first deployment earns the next one.