Sales teams spend a surprising amount of time on tasks that are not selling: updating records, writing follow-ups, searching for contact details, and preparing for calls. A wave of startups, many founded by engineers from large tech companies, now sells software that uses AI to take over parts of that work. The pitch is more revenue from the same team. Some of it is real, and some of it is optimism. Here is how to separate the two.
What revenue software typically automates
Most products in this group cover a handful of jobs.
- Data entry. Calls and emails are summarised and logged to the customer record without manual typing.
- Lead research. The tool gathers public information about a company and suggests who to contact.
- Drafting outreach. It writes first versions of emails and follow-ups that a person edits.
- Call preparation and review. It prepares a brief before a meeting and analyses recordings afterwards.
- Forecasting. It estimates which deals are likely to close and flags those that have stalled.
None of these replaces a salesperson. They remove friction so that people spend more time with customers.
Where it helps most
Teams with messy data benefit first. If your records are incomplete because people dislike typing notes, automatic logging improves visibility for managers and makes forecasts more reliable. Teams that send many similar emails also save time with drafting help, as long as someone reviews the output.
New hires can ramp up faster if the tool surfaces call summaries and suggested next steps, because they learn from patterns across the team.
Where it falls short
- Generic messages. Automated outreach can sound alike. Buyers receive many such emails, and bland ones get ignored.
- Accuracy. Summaries can miss nuance, and research can be wrong or out of date. Treat output as a draft.
- Forecast confidence. A model’s prediction is only as good as the history it learned from. A new product or a market shift can make it unreliable.
- Privacy and consent. Recording calls and collecting personal data brings legal duties in many places, such as notifying participants.
- Over-automation. Sending more messages is not the same as selling more. Volume without relevance can harm your brand and trigger spam filters.
How to run a fair trial
Choose one narrow outcome, such as hours saved on admin or the reply rate to outreach. Pick a pilot group and a comparison group of similar size. Run both for at least a month, or a full sales cycle if that is practical. Measure the same numbers for each group and look at the difference.
Include quality checks: have a manager review a sample of drafted emails and summaries each week and score accuracy.
Questions to ask the vendor
1. Where is our data stored, and is it used to train models for other customers? 2. What happens if the AI makes a mistake in a customer-facing message? 3. Which CRM and email systems does it connect with, and how deep is the connection? 4. How do we export our data if we leave? 5. What does pricing look like as the team grows?
Change management matters
New tools fail when people feel watched or inconvenienced. Explain what the software records and why. Let the team shape the settings, such as which fields are filled in automatically. Celebrate small wins, like a rep who saved an hour a day, to build buy-in.
Managers should resist using call analytics as a punishment tool. Use it for coaching, with specific examples of good calls to copy.
Cost versus value
Calculate the value of saved time in a simple way. If each rep saves three hours per week on admin, multiply by the number of reps and the loaded hourly cost. Compare the sum with the subscription. Then consider whether those hours turn into selling time, which is the real gain.
Bottom line
AI revenue software can remove drudgery and improve data quality, which helps teams sell more. It does not replace judgement, relationship building, or sound targeting. Test it on a narrow goal, keep humans in the loop for anything customers see, and judge the result on your own numbers.
What a good pilot looks like
A pilot should run for a single sales cycle with one clear question, such as whether drafted follow-ups raise reply rates by a set margin. Fix the rules in advance: who is in each group, which metric counts, and what result would make you buy. Write the answer down before looking at the numbers. This prevents the common trap of finding a flattering statistic after the fact and calling it success.
Further reading
Related on this site: Stealth AI Models: Why Labs Test Anonymous Models in Public and Why AI Labs Keep Hiring Each Other’s Researchers. For general background, see the Wikipedia overview.
