
Five questions the board asks about AI
Sooner or later the board asks five questions about your AI agents: how many do we have, are they under control, what are we spending, is it going up, and is it worth it?
- Today the answers come from last month's invoice and a spreadsheet someone stitched together.
- By the time it reaches the board, it is already out of date.
- Behind all five sits one bigger question: should the program go on? Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.¹
When the board asks what your agents are doing, last month's invoice is not an answer.
Every request counted, priced and recorded
Argent Strategic Intelligence sits between your AI agents and the AI models they call, such as ChatGPT, Claude or Gemini. Every request through it is counted, priced and decided, on a signed record.
- Counted. Each request belongs to a registered agent, and every agent has a named human owner, so every number has a name behind it.
- Priced. Argent reads the tokens from each answer, including tokens read from and written to the cache, and prices them from its own price catalog.
- Decided. Your policy allows or refuses each request, and the decision is written to the signed record.
- Read from the record. The console's figures are read from that record, so the board's numbers and the evidence behind them come from one place.
One console, five answers
Each of the board's questions has a place in the Argent console.
| The board asks | Where Argent answers | What it shows |
|---|---|---|
| How many agents? | Home | Working now: the agents that sent a request in the last two minutes, against how many are active. Plus the last 24 hours of requests and cost, each against the 24 hours before, and tokens with the share read from cache. Spend by owner: the people behind them, ranked by what their agents spend. |
| Are they under control? | Reports | Security posture for the last 30 days: checks that need attention and the share of requests Argent refused, each against the period before, and how many agents have no live owner. |
| What are we spending? | Live traffic, cost view | Spend adds up as each request is priced, agent by agent, with each agent's share. |
| Is it going up? | Home, Spend and Budgets | The last 30 days against the 30 before. Each budget's spend against the cap you set, with a marker at 80%. |
| Is it worth it? | Work items | Each project's spend against its budget, booked when an agent says which project it is working on. |
How it works, step by step
No new reporting project. The answers build up as your agents work.
- Route your agents through Argent. Each agent is registered with its own key and a named human owner, and its requests pass through the Argent gateway on the way to the model.
- Every request is priced. Argent reads the tokens from the answer, prices them, and writes the decision and the cost to the signed record.
- Set budgets where the money sits. Put budgets on the company, organizations, departments, people and agents, and on the projects in Work items.
- Agents say what they are working on. When an agent's session names its project, feature or story, Argent books that session's priced spend there and rolls it up to the project, without rewriting the record.
- The console keeps up. Home refreshes on its own. Live traffic's cost view adds up spend as each request is priced, so you can check it at any point.
- Before the board meets. Open Reports, read each check against the period before, and use EXPORT VIEW AS PDF.
What the operator sees in the console
Real screens, real labels. Every figure on them is read from traffic that runs through Argent.
| Where | What it shows |
|---|---|
| Home | Tiles for Working now, Requests 24h, Tokens 24h, Cost 24h and Top budget. Cards for Tokens used (fresh input, output, cache write, cache read), Estimated cost and Budgets. Ranges of 1h, 24h or 1w. |
| Home, added cards | Model mix (each model's share of tokens, cost or requests), Spend (30 days against the 30 before) and Spend by owner. Each person lays out their own board with Edit layout. |
| Live traffic, cost view | Cumulative spend with a sparkline, over a table of Agent, ~Spend (estimated) and Share, fed by the same live stream as the decisions table. |
| Reports | SECURITY POSTURE: CHECKS, REFUSED REQUESTS and AGENTS WITHOUT OWNER, a table of checks this window against the prior one, and EXPORT VIEW AS PDF. |
| Work items | Projects, features and stories. Items with a budget show a meter and their spend: within cap, near limit, over cap or paused at cap. |
| Who sees what | Each person sees their own part of the company. An Auditor reads Home and Reports without the money. Branch administrators see only their own unit's budgets and spend. |
Board oversight is already written down
For many companies, board oversight of cybersecurity risk is already in the rules they answer to: some must describe it, some must do it.
- US public companies, SEC Regulation S-K Item 106(c): they must "describe the board of directors' oversight of risks from cybersecurity threats," and describe management's role in assessing and managing those risks.²
- New York financial services, NYDFS 23 NYCRR 500.4: the CISO "shall report in writing at least annually to the senior governing body on the covered entity's cybersecurity program," and that body must "exercise oversight of the covered entity's cybersecurity risk management."³
Where AI agents are part of that program, Argent's Reports page gives the report its AI figures: agents without an owner, refused requests, and each check against the period before where an earlier reading exists, exported as a PDF.
Where it fits in your program
Built for the people who answer to the board for AI: CISOs, CIOs, CFOs and the teams that run the agents.
- Goals: Budget and token spend, See all AI in your company, Agent identity and ownership, Audit and accountability and Compliance evidence. See all goals.
- Industries: Financial services, Insurance, Healthcare, Public sector and Retail, or see all industries.
- Use cases: Rolling out copilots and chat safely, Securing AI coding agents and Governing customer-facing AI assistants.
- Pairs with: AI Spend Caps to stop spend at the cap, Agent Work Tracking to book spend to projects, Model Control to decide which models do the work, Identity Tethering so every agent has a named owner, and One-Click Audit Evidence to prove it. See all capabilities.
- Deployment: air-gapped on premises, private cloud or your own public cloud VPC; compare options.