LLMs don’t solve communication problems. Yes, an AI notetaker will create wonderful meeting summaries and with ChatGPT every email will seemingly be spot on. But this doesn’t solve the problem of mediocre communication skills. I wouldn’t count myself as a great communicator, but I have learned a lot over the past few years through practice. Still, I surely have many blind spots when it comes to reflection on professional interactions. What if a little machine could help me with that?

In “The Start-up CTO’s Handbook”, Zach Gold outlines a typical meeting schedule for a software engineer:

• One hour: 1:1 with manager
• Two and a half hours: daily thirty-minute standup meetings
• Two hours: average time spent in other agile ceremonies (sprint planning, retrospectives, etc.)
• Four to eight hours: reviewing others code
• Four hours: email/chat communication

In total, that's about thirteen to seventeen hours of the week used up for meetings and communication. If you add another few hours on top of that for time spent context switching and unplanned miscellaneous interruptions, quickly you're looking at at best half of a forty-hour work week available for actual focus time. If you're not careful about when meetings are scheduled, then not only will your engineers have only twenty hours left for their core tasks, but also they won't have them in contiguous blocks, further reducing productivity.

With meetings being such an integral part of work, possibly half of your working time (!), it would make sense for that to be half of the assessment criteria when hiring a new employee. But if an engineer went to school for 4 years to get a technical degree, he definitely didn’t spend half that time learning communication. Traditional education certainly teaches us to become better communicators through group projects, presentations, and other assignments, but it is simply not a focus issue.

Simply taking less meetings is also not a feasible option, as collaborative work and alignment with other employees are core for developing a proper business. So how is technology helping us currently?

AI notetakers have been heralded as the solution to many of the problems of meetings. You can focus on talking and have a perfect summary in the end. Everyone is supposedly on the same page. Every meeting ends with clear action items. But that doesn’t make the meeting run significantly more effective. Yes, you save time taking notes, but conflict resolution and the way you communicate yourself in the first place are still a massive issue.

It is a psychological problem, to understand the shortcomings of someone’s communication. So how do we solve that? Can technology help us gain insight into our own communication and that of our coworkers? Could it maybe even help us uncover biases in how we do communicate towards whom? 

What lies beyond the AI notetaker and AI email writing feedback? What tool could help us change our own behaviors and develop skills instead of relying on AI to communicate anything ourselves? 

Let's attempt a solution

I will define the core problem as follows: Reduce the time spent on meetings and communication by improving communication skills. I will further boil that down to effectively communicating an idea with little fluff and being receptive to other people’s reactions. 

Not another meeting AI

Meeting agents are optimized to summarize and extract to-dos. They fundamentally aren’t coaching tools that might give you feedback on what you said. They only look at the transcript for information. The goal of this solution is quite the opposite. Whilst it is important to look at what was said to understand how well it was said, the latter is also dependent on how you react to questions, doubts or more general reactions to your statements. 

Nonetheless, the existing solutions have some interesting ideas related to the issue:

• Fathom offers an "excessive speaking alert" – Funny but also condescending.
• fireflies.ai has quite the interesting solution called conversational intelligence.
• Zoominfo also offers some conversational intelligence.
• Gong has a solution to semantically analyze prospect clients.
• Covin tries to sell you on a similar idea.

Specifications and scope

The solution should go beyond the scheduled meeting, look at your written communication and combine those insights with any impromptu meeting or qick call. Of course nobody wants someone to be listening in all the time, but those short interactions hold lots of value when it comes to understanding professional relationships.

Clear guardrails on when the solution would listen in and when not, controlled by the employee, would be necessary. Also, transparency on what information is processed and how. All this should definitely run on-device. Not even your employer should have access to the raw transcripts of your day-to-day interactions. 

For someone to genuinely take feedback and implement it on their next conversation, you shouldn’t be babied into every new interaction. Telling the user to be extra understanding or gentle with a certain team member, reminding them of little things like a birthday to politely inquire about – all those things should remain human responsibilities.

Yes, it might make the product more appealing, but probably create new co-dependences and not create a more able human. Remembering those details and transferring experience to future interactions are part of being a good communicator, someone you like being around, and simply a decent human being.  Yes, clear feedback implies giving such clear directions, but it should always be retrospective.

Capturing written interactions

Let us look at the more structured part of communication first. What could you check emails, texts and slacks for?

• Passive aggressiveness: Is the tone appropriate? In which contexts does it break etiquette?
• Bias towards individuals: Is the person, maybe a manager, equally engaging with and supportive of all team members? Do they unconsciously single someone out?
• General style of communication: What weaknesses and strengths impact your work? How could you continue developing your unique style and fix shortcomings?
• Breakdowns: Who was actually at fault when bad communication leads to problems? Explanatory and neutral instead of berating and definite sounding feedback could make the tool an active aide in conflict resolution.

Silence is gold

Figuring out information summary of the spoken word is easy, or at least a solved problem. So what about non-verbal communication and tone? What extra information could be extrapolated by monitoring your webcam feed in video conferences?

Feedback and insights could include:

• Tells: Do certain behaviours unintentionally signal your interal state? Do you start excessively drinking water when nervous? Do you start looking to the side instead of at the screen when you get bored?
• Gestures: Do you use rich body language? Are there movements typical to your communication style?
• Comparison: Is your verbal communication as concise as your written one? Or do you tend to put a lot of fluff into your statements? Might you be too straight to the point?
• Agreement: How do others get you to agree? Might you even get tricked sometimes? What about the other way around?
• Attention: Did you listen to everyone else properly?
• Conversational styles: Do you understand other people and adjust to their preferred way of communication?

Most of this necessitates for both parties to consent to using the solution or having an instance running of their own. So for this to get adopted, it should first be sold as a B2B product into organizations.

Coach or friend?

The cool thing right now are coaches who interactively help you go through life as a friend/buddy/assistant. I would like it to be more muted, formal and generally neutral. It could be an abstract thing like ChatGPT is, and have a more clearly defined “professional relationship” with you.

It shouldn’t constantly inject itself into your daily work, but maybe pop up when things go really bad or at the end/beginning of the workday. Otherwise only give you stats when you refer to them.

Growing and learning

In writing this, I also came up with a few generally interesting features, that might help the solution generate value and be better marketable:

• Attention modules:
  • How good are you at negotiation? (Look at all details relevant to negotiation, from relevant arguments to leaning into information unveiled by the “opponent”)
  • How good are you at interviews?
  • How good are you at ad-hoc crisis reactions?

--> These all could be separate sections you can buy individually, or as a an add-on.

• Anti-Scroll: A simple alarm noise when you get distracted from a meeting or a warning of some sort.
• HR's best friend: Ability to share your reports with coworkers for feedback and their opinion on if it even makes sense.
• Adaptability: It should listen to your take and learn from how you perceived a situation. The tool could be two-sided and not only give but also take feedback to phrase and clarify future analysis.
• Comparative reflection prompts: “Last week you really struggled to convince Harry to get on your side with issue XY. One factor was how you didn’t accommodate for their preferences. Today you managed way better. What changed?” Contextual awareness would allow for supporting active reflection and recall of past interactions, reinforcing the feedback by directly combining it with past experiences.

Building for high churn.

Again, a live alert when typing out a passive-aggressive email is not the main goal here. In my opinion, that would hinder the learning experience. Decoupling feedback from the situation is definitely a risk and might reduce effectiveness, but would also reduce reliance.

This might be the result then: A weekly recap of your state of communication, repeating top learnings and growing points. Combined with valuable prompts for you to write out your thoughts and reflections. This would be enough in my opinion if you were to dedicate time to becoming a better communicator. One hour to go through last week’s communication, see specifically were things went wrong, and formulating insights for the future. You could even use the transcripts and specifically revisit situations where a meeting might have completely gotten derailed by a comment or when an email prompted a very spicy answer.

Grammarly already solves the problem of effective text communication. It is very good at simply making what you say, sound better. But it is not contextually aware beyond individual messages, and also not educational. Truth is, most people simply want the better result without putting in extra work. My critique might become more clear in this question:

How many people do you know that got rid of Grammarly after starting to use it?

My outlined solution would be designed to make itself redundant.

Best

Friederich