The agent team: your chief of staff holds the mandate, briefs the roster, and the agents work it together

Most procurement teams do not lack agents. They lack coordination. A researcher pulls a benchmark, a contract reader marks a clause, someone drafts a counter, and three days later two of those outputs quietly contradict each other. Nobody planned the sequence, nobody owned the handoff, and the reconciliation happens in a hallway conversation that never reaches the record. We built a stage where your team works in the open to fix the visibility problem. This update fixes the coordination problem. AI agents now has a second room, the plan, where a stated quarter becomes a briefed, gated, reconciled program of work.

The problem: a roster without a mandate

Six specialist agents and ten inbox agents can do a great deal of work. What they could not do until now was agree on what the quarter was for. You would open one, ask it a question, get an answer, open the next, ask a related question, and hold the whole plan in your own head. The head is where plans decay. Dates slip because nobody owned the renewal clock. Findings collide because two agents read the same contract differently and neither knew the other existed. The mandate lived in your inbox, not in the system, so the system could never enforce it.

The plan puts the mandate somewhere durable. You say what the quarter is for, for example "take three per cent out of the observability stack before the two September renewals co-term," and your chief of staff drafts a plan from your actual estate. Not a template. Your renewals, your contracts, your spend, read off the board that the morning open, the stage, the Monday brief and the Vera bot all already share.

What the chief of staff drafts

The draft is specific. Each workstream carries steps, each step carries an owner from the roster, and each date is pinned to a real clock, your renewal dates, not an arbitrary sprint. Wherever a step would send an email, sign a document or spend money, the plan inserts a gate. A gate is not a suggestion. It is a stop that waits for you. This is the same discipline as the renewal plan, where you state what you want and approve the steps, extended from a single renewal to the whole quarter and the whole roster.

Because the draft reads from your estate, it inherits your constraints. If two renewals are about to co-term, the plan can flag it rather than blindly align them, which matters given the co-term trap where aligning dates quietly hands the vendor leverage. If a spend list already ranks your moves, the plan can pull from that savings plan so the quarter starts from evidence, not ambition.

How the team works it

Approve the plan once and the team works it. Each owner is briefed in its own conversation. It acknowledges the brief or asks back if the brief is thin, does the work, reports the result to the ledger, and hands its finding to the step that was waiting on it. That last part is the coordination that used to happen in your head. The step that needed the benchmark receives the benchmark, cited, without you carrying it across.

When two readings differ, they are reconciled in the open. If the contract reader and the researcher disagree on the true unit price, the disagreement is surfaced on the board and resolved on the record, not buried in whichever output you happened to open last. This is the same principle behind reports over answers. A finding that can be inspected and traced beats a confident sentence you cannot audit.

What stays with you

Send, sign, spend, hire and stop stay with you. They live on the card and take one press. The agents can do everything up to the gate, assemble the counter, draft the message, line up the signature, and then they wait. You approve on the card or you press stop. Nothing crosses a gate without you, and the audit log records each act against the version of the plan you approved. If you change the plan, the log shows which version was live when each act happened.

Everything reads the same board. The morning open, the stage, the Monday brief and the Vera bot in Slack and Teams all draw from one plan. There is no reconciliation between what the bot told you on Friday and what the dashboard shows on Monday, because they are the same object. That single source is what makes the audit log meaningful rather than decorative.

Honest limits

Be clear about what this does not do. The plan is only as good as your estate. If your contracts are not loaded, or your renewal dates are stale, the draft will be thin and you will spend your forty minutes fixing inputs rather than approving a plan. The chief of staff drafts from what it can see.

The agents do not negotiate for you past a gate. They prepare, they assemble, they wait. When a vendor's own agent is on the other side, the exchange follows the open Agent Negotiation Protocol, but the decision to send, sign or spend is still a human press. That is deliberate. A gate you can override is a gate. A gate the system removes to save you time is a risk.

And reconciliation surfaces disagreement, it does not always settle it. When two agents read a clause differently and both readings are defensible, the plan shows you both and asks you to choose. That is the right behaviour, but it is your judgement being asked for, not spared. The plan reduces the labour of coordination. It does not remove the responsibility of the buyer. If you want that responsibility grounded in evidence before the quarter even starts, begin with the kickoff interview and let the plan build on what it learns.

Want to be updated when major licensing and pricing changes land? One analyst brief a week: the price rises, metric changes and audit campaigns that move software costs. Work email only.