Define What You Want to Improve
Before diving in and starting any development, understand what you are trying to improve. Often businesses get carried away and start looking at features and forget what the issue is. However, taking your time to plan can help reduce any costly missteps.
When looking into using AI agents for sales automation, think about what issues your sales team face regularly. For example, if you’re growing your B2B business, you might have inconsistent follow-up, incomplete customer data, or poor communication between sales reps. Ask yourself what would be made easier by solving this issue.
As different people will have different expectations, talk to each department and understand what they expect. Sales representatives, managers and revenue ops often have different views.
Sales managers may see that their sales reps are inconsistently using opportunity stages in their CRM. This can cause confusion when it comes to reporting.
Ask employees how they go about researching their prospect, updating their customer data and what they do next. Try to get real examples.
Creating a simple problem statement will be a useful reference throughout your project. This will help keep everyone focused when new feature requests come up.
Talk to the People Who Experience the Problem
Once you know the main issue within your business, you’ll need to understand how it impacts those involved. Whilst a planning document will outline what your business wants, talking to the people who use it will show you what they really need.
For example, your account executive may be spending a few minutes each time looking through old emails, CRM data and asking colleagues for information to prepare for a meeting with their prospect. Whilst this seems simple for management, it can cause frustration for the employee.
When thinking about using AI agents for sales automation, this can help you understand what features to develop.
Instead of asking employees what features they would like, ask them to think about a recent situation that caused them issues. What went wrong? What information did they lack? What part of the process took the longest? How did they overcome it?
Looking at examples of their work can also help. Taking a look at some anonymised customer data, an internal checklist, example email or handover note can give you more detail.
Speak to your sales managers and revenue ops too. They can help to show any differences between what is supposed to happen and what they are doing.
Having a conversation with someone who isn’t convinced automation is needed can also help you see any risks or issues.
Understand How Your Current Sales Process Works
Before you start automating anything, take a look at how you currently work. Having a good understanding of how you currently run your processes will help you understand where you can add real value with an AI agent.
Start by mapping out your current process. Where is information flowing? From your CRM, calendar, approved email system and product catalogue. What manual processes are happening?
You may already be using automation within your email, but if your contact data is outdated and incomplete this won’t help. You need to address the data first.
Remember that if your process is inefficient, it will remain inefficient once automated.
Map out the full process of important customer data. Where does the data come from? Who updates it? Who does it get passed onto? Do employees trust the data? Are they having to manually update it? Is the data being duplicated?
Sometimes duplication isn’t the issue. It could be that there isn’t clear ownership, missing information or lack of standard processes.
For any important handover, there should be a clear starting point, owner and definition of done. This will give your development team a much better understanding of what to connect and improve.
Start With a Small Experiment That Delivers Value
One of the easiest ways to lose control of your development is to try to do too much in your first version. Start by creating a small experiment that will answer a key business question.
Maybe You Want to understand if your AI agent can help with researching your prospects, updating CRM data and suggesting the right next steps, whilst reducing manual effort.
Before you start, think about how you’ll measure success. What will show that you need to improve further?
If you’re sending three similar follow-up emails to your prospect due to the different tools not knowing what has been communicated recently, a small experiment can help you understand if improving the integration will help.
Try to keep your experiment small and include both simple and complex scenarios. How will the system handle complete customer data, outdated contact details and opted-out customers?
Make sure there is a clear end point. If the experiment isn’t providing reliable results or causing new issues, you’ll want to have the flexibility to rethink your approach.
The aim of the experiment is to learn something useful. A small experiment with careful planning will give you more meaningful insights than a large project based on assumptions.
Make Important Decisions and Trade-Offs Clear
Planning will help you know what features to include in your first release.
Whilst a feature may sound useful in a meeting, think about how much value it will bring. Will it increase development time, complexity or maintenance?
Even if your AI agent updates CRM data correctly, the email draft may still be attached to the wrong customer. This shows the importance of ensuring your systems are properly integrated.
When planning your AI agents for sales automation, consider the features against the requirements (CRM integration, data cleaning, approvals and monitoring).
Write down any important decisions. This will help avoid them getting lost in your meeting notes. Explain why a feature has been included, delayed or excluded.
This will also help you identify any core features and nice to have features. Maybe researching your prospect, updating CRM data accurately and suggesting the right next steps are core features, but you can delay adding more complex workflow options.
Delaying a feature doesn’t mean it isn’t useful. It just means You Want to gather more evidence before investing extra time.
Talk to those who will be using and maintaining the system to review your decisions. Highlighting where compromises have been made will help everyone understand what direction your project is going in and reduce any unnecessary arguments.
Build Regular Feedback Into the Planning Process
Never become too attached to your original design. Those using the system may see issues that you and your managers didn’t consider.
For example, a customer may unsubscribe from a marketing campaign on a Friday, but then get an email sent to them on a Monday. This is a clear example of where processes aren’t integrated and has created a bad customer experience.
This shows the importance of having feedback loops.
Ask your sales reps what they think. Are there any manual workarounds, confusing processes or extra help needed?
Speak to those who weren’t involved in the initial planning. They may notice confusing instructions or missing steps as they’re seeing the process for the first time.
Combine this with any metrics you have. Are you seeing good responses, conversion rates from meetings and accurate activity within your CRM? If they’re showing one thing but your employees are saying something different, investigate further.
Having feedback loops will help your business identify any real issues and improve based on how it works in practice.
Turn Your Research Into a Clear Development Plan
Your discovery should leave you with a clear plan that your development team can follow. Ensure you’ve captured any key insights, decisions, owners and questions.
For example, your AI sales agent could be used to research your prospect, gather the right information and create a personalised message. The sales rep will then review the content to decide if they want to send it.
This means you’re automating the repetitive task, but leaving the important customer interaction to a human.
A useful development brief should include your current process, what issue you’re trying to solve, what you hope to achieve, what features you need in your first release, any risks, key assumptions and who owns what.
What will you measure in your first production output and who will review it? If there isn’t clear ownership, even if your solution is well designed, it won’t be easy to track and improve.
Your roadmap should give you direction but shouldn’t lock you into making decisions. As you learn more during testing and use, priorities may change. This is natural when creating a useful product.
Overall, great planning for your AI sales agents will come down to understanding your issue, listening to those involved and making decisions based on evidence. Having a clear discovery process will give your development team a solid foundation and help your business move forward with what they originally wanted to achieve.

Top comments (0)