Learnings from AI Coach Beta

Intro

I’m super excited to announce that Wonderway Coach is now coming out of beta and will officially launch in October! As we prepare for the launch, I wanted to take some time to reflect on the journey and share some of the key learnings with all of the people that have shared feedback along the way.

‍

The vision

Imagine a world where every sales person could have access to their own personal sales coach. This coach would not only know all of the sales fundamentals, it would understand you personally, knowing your individual strengths and weaknesses, what motivates you and how you like to get feedback. The coach is available 24/7 and knows exactly what you need to hear, when you need to hear it. No more waiting on your manager or wading through thousands of articles and videos trying to find the slice of information you need. In this world, everyone has a fair chance. Potential isn’t hindered by a lack of resources or guidance.

‍

Where it came from

When the AI revolution started almost a year ago, this was the vision we had. After spending the last 5 years building sales training products, we knew that sales coaching was broken. All of our customers were using conversation intelligence tools to record their sales calls but nobody had time to watch the recordings. They knew that the answers to some of their most important GTM questions were locked away in the calls but they couldn’t access them as it just took too much time to watch the calls and categorise the feedback.

‍

The limitations of conversation intelligence tools

What surprised me is that everyone seemed to have been fooled by a common misperception. Conversation intelligence tools don't coach—they merely facilitate the process for human managers. If managers don’t embrace them, their value effectively diminishes to zero.

Many of our customers that didn’t understand that when they bought the tools and were surprised by how basic the keyword analytics were. The reality is that conversation intelligence tools can only give you insights into the quantity of interactions but provide zero insight into the quality. In other words, they will tell you if you asked questions (duh?), but not if the questions you asked were the right ones. Example:

What this means is that if a sales person wants to actually improve, they still needed to wait for their manager to listen the call and provide qualitative feedback. And so started the conflict between management who tried to force managers to listen to calls and fill in scorecards to figure our what was going on while team leads resisted as they didn’t have the time.

‍

The importance of sales coaching

By now, I think everyone knows how important sales coaching. But for those on the fence, here are some facts from CSO insights (the research division of Miller-Heimann):

1. Companies with dynamic coaching programs achieve 27% higher win rates and 16.7% higher annual revenue growth

2. 60% of sales people are more likely to leave their job if their manager is a poor coach

3. But 47% of managers spend <30 mins a week coaching

‍

The challenges with human coaching

So if everyone knows how important coaching is, why don’t people do it? Well the brutal truth is that sales coaching is important but rarely urgent. Sales managers are super busy and it always get’s pushed when more urgent things come up. And even if you’re one of those sales managers with an iron will to do what’s right you’re still in trouble:

1. Managers almost never get taught how to to coach

2. There are just way too many calls to watch. Most sales managers have a team of 8 reps. If every rep is doing dozens of calls a day/week, its impossible to managers to watch them all

3. Feedback isn’t in real time. Takes time for managers to listen to calls and share feedback. By that time, the coaching is less effective as it was so far in the past and people forget.

Let’s face it. Coaching just wasn’t meant to be done by people. Luckily we are now at the point where we can stop just using AI to write emails and start leveraging it for one of the most powerful applications it was meant for - sales coaching.

‍

The opportunity for AI sales coaches

What makes us different to existing conversational intelligence tools is that we use large language models to analyze the calls rather than keyword analysis. This means we understand the full context of the call and can provide human-like feedback just as good as a real manager. And it couldn’t have come at a better time:

1. Just as the world was trying to figure out how to do more with less, there’s now a way to save managers countless hours listening to calls.

2. Just as sales reps were expected to work from home - putting them further away from their managers than ever before, there’s now a way to get the coaching and feedback they need wherever they are.

3. Just as most companies were changing their go to market motions, there’s now a way to see how closely the playbook is being following and make automatic suggestions for how it could be improved.

‍

Key learnings from our beta

Since launching the beta version in May, we have had thousands of people test the product and share feedback. These are the key things we have figured out along the way:

1. Customisation is key: Every company has its own playbook. Its essential that the scorecards can be adjusted to account for the different processes that each company has. Despite adding 15 different scorecards to the system for all of the key sales methodologies (SPIN, SPICED, MEDDIC etc.) we see that every successful company eventually creates their own scorecard to reflect their process to get the most out of the tool.

2. Training the model increases performance: Completely untrained the coach is 80-90% accurate when it comes to identifying the key gaps that sales reps need to improve (what you get with the product as part of the free trial out of the box). Once its trained on the companies data it’s 90-100% accurate (what we offer to paying customers).

3. Different scorecards for different types of calls: This was one of the biggest technical challenges to solve. The coach can now distinguish between different call types (e.g. cold calls, discovery calls, demos, negotiations, implementations etc.) to apply the right scorecard.

4. Sales reps want feedback on every call, managers want the aggregate: While sales reps love getting an instant summary and feedback on each call, its overwhelming for managers. They prefer to just look at the insight once a week with their team members to see what are the main areas for improvement and if things are improving over time.

5. Build trust with examples: We provide quotes from the transcript to back up all of the recommendations we make and make it easy to inspect the data so you can see where the feedback is coming from.

6. Transcription quality is paramount: After testing transcripts from all of the major conferencing tools, diallers and conversational intelligence platforms, we found that none of their transcripts were good enough to provide accurate performance feedback. We now use OpenAI’s Whisper to transcribe the calls which is in a league of its own in terms of quality and the number of languages it supports.

‍

Key successes

Along the way here are some of the key successes we have had:

1. Increased conversion rates by 39% for one customers inbound calls by identifying best practices from their top performers and applying across the rest of the team.

2. Achieved 94% accuracy in coaching results for one of our most recent triallists with the untrained model straight out of the box.

3. One customer found that the top performers in the company were actually the least likely to follow the playbook. Lead to a complete overhaul of the sales process to re-align the team around what the top performers were doing.

4. Helped one customer to optimise their playbook by identifying the most common objections. They re-wrote the discovery script to pre-emptively address the objections before they were even raised.

5. Signed on 3 of the top sales trainers in the world as advisors: We’re working together with John Barrows, Richard Harris and Kevin Gaither to help us build SLICE: our sales-led intelligence coaching engine to build their sales training and coaching experience directly into the product.

‍

Future developments

• Chat with your data: We believe the future of analytics interfaces will be chat. Rather building 100’s of dashboards, we are implementing a chat interface that will allow sales reps and managers to interact with their data similar to ChatGPT. We have already implemented the first version which allows managers to ask questions like:

1. What separates top performers?;

2. How can we optimise our sales process? and;

3. What changes could we make to the pitch and product to improve product-market fit? We will be more power here in the coming months to produce even deeper insights.

4. Modular LLM: We believe that in the future, large enterprises will have their own LLM. We will allow companies to switch use their own LLM to power the analysis.

‍

How the product works - try it out

It’s completely free to try out. Sign up for a free account here and we will analyse the last 5 calls for each person on your team to give you immediate insights into what is going well and what is going wrong on your calls. Here’s the steps:

1. Connect it to your calling tool (e.g. Zoom)

2. Choose which calls you would like to analyse (e.g. cold call, discovery etc.)

3. Choose a scorecard or build your own (e.g. SPIN, SPICED, MEDDIC etc.)

4. Add people from your team

5. We will analyze 20 sales calls for free