Human-in-the-loop means a person decides at the points of a process that matter, while the agent prepares and proposes. Where to put the human and why.
Human-in-the-loop
Human-in-the-loop is the principle that a person stays in control of what an AI system does, at the points that matter. The term became widely known in machine learning, where people correct a model’s outputs to train it. It now extends to AI agents, which act with real autonomy: the human no longer just trains the system, but frames what it does. In automated and agentic processes, the agent researches, analyses, acts and, above all, proposes. At the critical steps, a person validates the outcome, overrides or takes over, and every step goes on record.
Why human-in-the-loop matters now
The loop has moved. In machine learning, the human sat in the training loop, improving the model. With agents that read, draft, compare and act, the human sits in the decision loop, governing what the agent does. The question that decides whether agents reach production is not what they can do. It is who answers for what they did. Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027, and names inadequate risk controls among the three causes. Regulators ask the same question: for high-risk systems, the EU AI Act requires that a person can understand, intervene and override. The people expected to use these agents ask it too: at our September summit, 62% of the executives we surveyed wanted a human to validate every agent action. Human-in-the-loop answers all three, and it is what lets you put an agent on work that commits the company, like a discount or a contract.
Where to put the human in the process
Not every step needs a person. In our view, three places do. Approvals: a discount above threshold, non-standard contract terms, a purchase above a set amount, where the agent prepares the file and the manager decides. Exceptions, where most of the real work hides: a customer nobody recognises, a missing purchase order, a request that matches no rule. The agent flags them instead of guessing. And any action that commits the company to someone outside it, like an email to a customer, an invoice or a new user account. Everywhere else, let the process run. A manager asked to approve thirty routine steps a day stops reading them by the fifth.
Human-in-the-loop by design: the agent proposes, the human decides, the record keeps both
Agents can now do far more than draft. They research an account across systems, analyse a margin against comparable deals, compare a contract with a playbook, fill a record from several sources, create a document or a page, and act on their own within set limits. Two things never belong to the agent: the guardrails it runs within, and the outcome it produces. People set the first, a person owns the second.
Take a quote-to-cash deal at quarter end. The agent has prepared most of it: the quote, the account history, the standard contract, the brief for Finance and Customer Success. Two moments still need a person. The rep asks for 28% off, so the agent checks the margin, past discounts and comparable deals, and sends a recommendation to the COO, who owns anything between 20 and 40% (below, the sales director; above, the CEO). Then the client returns the contract with seven redlines. The agent applies the five that match Legal’s playbook. A lawyer reviews the other two, not the whole contract.
.png)
Then comes the part most teams skip: the record. Six months later, the CFO asks why that 28% discount went out. The record shows what the agent recommended and why the request went to the COO, then what the COO decided, and when. Across a few hundred deals, it also shows which recommendations people accept every time. That is where an agent can earn more room.
That is human-in-the-loop by design: before the first deal runs, the process model already says which steps belong to a person, to the system or to an agent, and what the COO sees when the discount lands. The COO spends minutes on the decision, not an hour rebuilding the file.
Human-in-the-loop and the EU AI Act
Regulation now names the principle. For high-risk AI systems, the EU AI Act sets two obligations that matter here: human oversight and record-keeping. Human oversight means a person can understand what the system does, intervene and override it. Record-keeping means the system logs its activity so that anyone can trace decisions afterwards. The NIST AI Risk Management Framework points the same way: accountability needs defined human roles and a record of what the system did. A process designed this way covers both, since the decision points give the oversight and the log gives the record. Not every business process falls under the high-risk category, and this page gives no legal advice. Still, the direction is clear. A CIO who can show where humans decide, and prove it with a record, has an easier conversation with auditors, regulators and the board. Building it in from day one costs far less than adding it later.
From validating every action to governed autonomy
Human-in-the-loop works like a dial. The autonomy literature names three positions: in the loop, the person approves before the action; on the loop, the system acts within rules and a person supervises and can take over; out of the loop, the system acts alone. Most organisations start at the cautious end, with a person validating every action the agent proposes, and they are right to: the first weeks surface the cases nobody wrote a rule for. After a month or two, the approval log tells its own story. A standard renewal quote approved forty times without a change is ready to move on the loop. A redline review that Legal still corrects one time in three stays where it is, and so does any action that commits the company to a customer. Steps move one at a time, never the whole process at once. That is governed autonomy: most quotes go through on their own, the odd one lands on a manager’s phone, and nobody wonders afterwards who approved what.
→ See Ofelia for operations and sales teams.
Ofelia and human-in-the-loop
Across the Ofelia suite, the principle is the same: agents work inside a frame people define, people decide, and every move lands in the audit trail. Where an agent acts depends on the product.
In Ofelia Agentic, the Ofelia assistant acts as an agent in two places. When someone asks a question in Slack or Microsoft Teams, Ofelia searches your approved documentation, answers and guides the next step, without leaving the conversation. The person stays in charge of what happens next. When a process owner builds a workflow, Ofelia proposes the model from existing documentation: the steps, the owners, the thresholds. The process owner reviews it, edits it in a few clicks, tests it in sandbox and publishes it. No developer, no IT ticket, and nothing goes live without that validation.
Once published, the workflow runs in Ofelia Workflow, and it is fully deterministic. It fixes the steps, their order, the owners and the thresholds. No agent decides the sequence. Human tasks go to the right person: the discount above policy routes by threshold to the sales director, the COO or the CEO, and the routing survives the next reorganisation. Automated system tasks update the record, file the document and send the notification. Some steps go to an agent: preparing the quote, checking credit and payment terms against policy, comparing redlines with Legal’s playbook. That is where human-in-the-loop matters most: the workflow waits for a person to validate the agent’s result before it moves on. In the deal above, Legal reads two clauses, not seven. When the client signs, nothing starts by itself: IT, Finance and Customer Success each get a card in Slack or Teams to open the provisioning ticket, send the invoice or book the kickoff, when they are ready.
In Bonita BPM, the same principle holds where the process needs its own application. Human tasks and approvals are modelled in BPMN 2.0 and assigned by role. The AI Agent Orchestrator delegates a step to an agent inside a process that keeps control: the agent acts within the model, the process decides what happens next, a person validates where the model says so. More than 15 years of running processes for more than 130 enterprise customers, from banks and ministries to insurers and retailers, taught us where that validation belongs.
AI speed, human control: on the 28% discount, the COO still decides, in minutes.
Download the Ofelia Agentic one-pager (PDF).
See Ofelia handle your actual workflows with Human-in-the-loop
Sources
European Union, AI Act, Article 14 (human oversight) and Article 12 (record-keeping). NIST, AI Risk Management Framework 1.0. Gartner, press release June 2025, over 40% of agentic AI projects will be cancelled by end of 2027. Ofelia Summit 2026 barometer (directional, internal).