How realistic is ROI from AI in sales? What the data shows

Hardly any vendor can resist impressive promises of returns. But how realistic ROI from AI in sales actually is can't be read from brochures, only from your own figures. This article explains why published values vary so much, how to check an ROI claim and, using a worked example with three scenarios, how far apart the results can be depending on the assumptions.

Why blanket ROI promises tell you little

When a vendor talks about a multiple of the investment, there is usually a real figure behind it, just not yours. Published values often come from the most successful projects. Businesses where little changed don't write success stories. On top of that come typical distortions:

• A weak starting point. If you previously missed half of all calls, you will improve a lot. If you were already well organised, you will gain far less.
• Revenue instead of contribution margin. An additional order worth 5,000 euros doesn't bring 5,000 euros of profit. Calculations based on revenue inflate the benefit.
• Unclear attribution. Would the customer have come without the AI too? Many presentations leave this question open.
• Missing costs. Internal time for maintenance, checks and coordination rarely appears on the cost side.
• Short periods. One strong month in peak season isn't an annual result.

That doesn't mean AI in sales brings no return. It only means you can't transfer other people's figures to your business.

What the data shows and what it doesn't

Industry reports and vendor studies show one thing above all: results vary widely. Well-executed projects with a clear bottleneck and enough enquiries can pay off handsomely. Projects without a proper baseline, without a CRM connection or without someone looking after maintenance often fall short of expectations. There is no reliable average that applies to a trade business just as much as to an online retailer.

If someone presents you with an ROI figure, these questions help you put it in context:

1. Who measured it, the vendor itself or an independent body?
2. What was it compared with: the situation before, or a control group?
3. Over what period, and did it include a season or special promotion?
4. Which costs are included, including setup and internal time?
5. Was the calculation based on revenue or contribution margin?
6. How similar is the example company to your business in size, industry and enquiry volume?

If more than two questions go unanswered, treat the figure as advertising, not as a basis for planning. A similar look is worthwhile with comparisons between people and machines; see what the conversion data on AI calling really shows.

Worked example: AI ROI in sales for a motorhome rental business

All values in this section are assumptions, not measurements. Suppose a motorhome rental business receives around 240 enquiries a month in season via phone, website and WhatsApp. About 40 of them currently go unanswered or are only answered days later, because the small team is busy with handovers and returns. A booking brings an average contribution margin of 600 euros.

From now on, an AI assistant answers all enquiries immediately, checks availability and makes reservations. For costs we assume a monthly licence of 399 euros, which corresponds to the Neurobots Basic plan, plus 1,299 euros setup spread over twelve months (around 108 euros) and four hours of internal maintenance at 40 euros. Together that comes to about 667 euros a month. The decisive question remains open: how many of the 40 previously lost enquiries now turn into bookings?

Scenario	Additional bookings	Additional contribution margin	ROI factor
Cautious	2	1,200 euros	around 0.8
Medium	5	3,000 euros	around 3.5
Optimistic	8	4,800 euros	around 6.2

The ROI factor is (benefit minus costs) divided by costs. The assumption about additional bookings alone shifts the result several times over. That is exactly why blanket promises are worth so little.

Looking at the whole year is even more important. In the off-season, our example might see only 60 enquiries, of which ten were previously left unanswered. That produces hardly a single additional booking, yet the costs keep running. If you only calculate the peak season, you overestimate the return. The article Measuring chatbot ROI explains the detailed formula with all cost items.

When ROI from AI in sales is realistic and when it isn't

The chances are good if several of these conditions apply: enquiries are regularly lost, an order brings a noticeable contribution margin, many enquiries arrive outside business hours, and the AI is connected to the calendar or CRM. You also need someone in the business who spot-checks conversations and adjusts answers.

It gets difficult with very few enquiries, with orders that have a low contribution margin, or when capacity is already fully used. A rental business whose vehicles are completely booked out in season won't gain a single additional booking from faster replies. At most it gains time in the office. Where AI still helps a rental business, for example with handovers and fleet management, is described in AI for rental companies.

How to check ROI in your own business

1. For four weeks, record enquiries, missed contacts, response time and deals. That is your baseline.
2. Work through three scenarios like the ones above in advance. If the cautious scenario doesn't come anywhere near paying off, be sceptical.
3. After launch, tag every enquiry that comes via the AI in the CRM.
4. Evaluate after three months, and for seasonal businesses again after a full year.

The article on the key KPIs for AI sales automation describes which metrics to keep an eye on. If you later have to present the results to shareholders or the bank, the article Measuring and communicating AI ROI will help.

Frequently asked questions

How long does it take for AI in sales to pay off?

There is no general answer. In the worked example, the setup costs would be recovered after the first month of the season in the medium scenario, while in the cautious scenario it takes much longer. Calculate with your own values.

Are vendors' ROI figures inherently unreliable?

No, but they usually describe selected cases. Ask about the method, the period and the basis of comparison. A reputable vendor can answer these questions or will say openly that it can't.

Should I include time savings in the ROI?

Yes, but only if the time saved is actually put to other use or costs disappear. Half an hour saved a day that simply evaporates in everyday work only improves the calculation on paper.

What if the ROI doesn't show after three months?

First check whether the right process was automated and whether the AI has access to the calendar and CRM. If that isn't the problem, the honest consequence is sometimes to end the project. That is a result too.

Conclusion

A good ROI from AI in sales is possible, but it doesn't happen automatically. The published data mainly shows a wide spread, and your result depends on enquiry volume, contribution margin and starting point. Calculate with three scenarios, over a full year and with all costs. If the cautious scenario holds up, the project is a sensible decision.

The page Digital rental assistant shows how an AI assistant handles enquiries and reservations at a rental business.