How Much Does an AI Agent Cost, and How Do You Calculate Its ROI?

“How much does it cost?” is the first question any business owner asks about an AI agent. It's a good question, but it's only useful if it's followed by a second one: “What does it bring in?” An agent at €49 a month can be a great deal; an agent at €9 can be useless.

This article breaks down the full cost of an AI agent, offers a method for calculating return on investment and helps you decide with numbers in hand. The worked examples are fictional cases meant to illustrate the method: replace them with your own data.

What an AI agent costs: three components#

1. The price of the tool#

This is the visible part. Depending on the solution, it can take several forms: a per-user subscription, usage-based billing (number of executions, operations or messages), a flat fee per agent, or custom development billed by a contractor.

On Future Agent, each agent has its own monthly price, set according to how demanding the task is and quoted by the architect before launch. You know what you'll pay before you commit. The public benchmarks are:

Most agents fall between €9 and €49 a month. On top of that, there's an optional plan that bundles several agents: Free (€0, you only pay for active agents), Pro (€29, 5 agents up to €9 included and 20% off additional ones) or Studio (€79, 12 agents up to €19 included and 35% off additional ones). Details are on the Pricing page.

2. Setup time#

This one is often overlooked. For an agent, it includes:

• the time to describe the task and answer the questions (one conversation);
• the time to connect your tools (a few clicks per tool);
• the time to approve the sample and adjust the instruction if needed.

With a conversational tool, that's anywhere from a few minutes to an hour depending on the task. With custom development, it can run to days of contractor time. With a workflow tool you build yourself, it depends on your technical comfort.

3. Supervision time#

An agent isn't “set and forget.” In the first few weeks, you read its summaries. After that, you glance at the log now and then and adjust an instruction when your business changes. Count this time honestly in your calculation: it's small, but it's real.

What an AI agent brings in#

Time given back#

This is the main gain and the easiest to measure. The question is simple: how much time do you spend on this task each month? Time it for a week or two rather than estimating from memory.

To put a value on that time, use the fully loaded hourly cost of the person doing the task (gross salary, employer contributions and overhead divided by hours worked) or, if it's you, the value of an hour you could spend selling or delivering.

Indirect gains#

Harder to quantify, but often more important:

• consistency: a quote always followed up on day 5 instead of “whenever someone remembers”;
• responsiveness: a client request acknowledged within minutes;
• fewer slip-ups: a contract deadline, a missing document or an unpaid invoice no longer falls through the cracks;
• team availability for higher-value work;
• quality of deliverables: reports always in the same format, complete and on time.

Don't inflate these gains to justify a decision. Note them separately as qualitative benefits, and only include what you can measure in the calculation.

The calculation, step by step#

Step 1: Measure the current time#

Monthly time = time per occurrence × number of occurrences per month.

Step 2: Estimate the share that's truly automated#

An agent doesn't always eliminate 100% of the time. If it prepares drafts you review, there's still approval time. Estimate the share of time that genuinely disappears.

Step 3: Value the gain#

Monthly gain = time saved (in hours) × fully loaded hourly cost.

Step 4: Calculate the full cost#

Monthly cost = agent price + share of the plan + valued supervision time.

Setup time is amortized over the first year: divide it by twelve and add it to the monthly cost if you want to be rigorous.

Step 5: Compare#

Net monthly gain = monthly gain − monthly cost. Return on investment = net gain ÷ cost, expressed as a percentage.

A complete (fictional) example#

A service company spends time following up on unanswered quotes. Here's the calculation, with assumptions to replace with yours.

In this example, the net gain is about €170 a month, a return of more than six times the full cost. And that doesn't count the quotes signed thanks to more consistent follow-ups, which would be tempting but unwise to put a number on without data.

The same reasoning can lead to an unfavorable result: a task that takes five minutes a month doesn't justify an agent, even a cheap one. That's the whole point of running the numbers.

Optimizing your budget with several agents#

Once you have several agents, the plan you choose matters. A few examples based on the public rules:

• Five agents at €9: €45 on the Free plan, €29 with the Pro plan, which includes all five.
• Five agents at €9 and one at €19: with Pro, the first five are included and the sixth gets 20% off, i.e. €15.20; total €44.20.
• Five agents at €19: €95 on Free, €79 with Studio, which includes up to 12 agents at that tier.

Your dashboard shows a breakdown of your bill and how much your plan saves you. You can switch plans at any time, prorated.

Project agents (websites)#

A project agent builds a hosted deliverable, such as a business website. Its cost works differently: it replaces a design-and-maintenance engagement. On Future Agent, the build and the first two revisions are free in preview; after that, hosting is monthly (for example €29 to €49 for a business website) and includes 10 revisions a month, with higher tiers if you need more.

To compare, put it next to the cost of having a contractor build the site, plus hosting and small changes billed one by one, over two or three years. Example template: Tradesperson website.

Pitfalls that skew the calculation#

• Overestimating time spent. Time it instead of guessing.
• Forgetting supervision. It's small but real, especially at first.
• Counting hypothetical gains. “This agent will get me two more clients a month” isn't data. Measure after three months.
• Comparing with zero. The alternative isn't “doing nothing”; it's often “doing it badly or late.” Note that among the qualitative gains.
• Ignoring duration. An agent that runs for three years pays off for three years. Custom development that breaks at the first tool change means paying for a new engagement.

A decision grid#

Before launching an agent, fill in these five lines:

1. Current monthly time, measured: ……
2. Realistic automated share: …… %
3. Fully loaded hourly cost: €……
4. Agent's quoted monthly price: €……
5. Monthly supervision time: ……

If the net gain is clearly positive, launch. If it's close to zero, bundle the task with another one, or wait until it grows. To find the tasks that pay off most, see our list of 25 tasks to hand off to an AI agent.

Measure afterwards#

Three months after launch, review:

• how much time you still spend on the task;
• the number of successful runs, visible in the dashboard;
• the corrections you had to make;
• the effects you observed (response times, follow-ups sent, slip-ups avoided).

This real data will replace your assumptions and tell you which agents to keep, adjust or stop. An agent you no longer need can be paused in one click.

In short#

• The cost of an AI agent = tool price + setup + supervision.
• The gain = time given back × fully loaded hourly cost, plus qualitative benefits noted separately.
• Run the numbers with measured figures, not optimistic estimates.
• With several agents, a plan can cut the bill significantly.
• Measure again after three months and keep the agents that prove their worth.