Mentorship and the art of actionable advice

ChatGPT's depiction of a robot mentoring a robot

One of my challenges with the transition to solo consulting is learning to deliver timely, actionable advice. It’s usually easy for me to identify many areas for improvement. Distilling a long list of “obvious” opportunities to the top items that would make a difference is harder. And the hardest thing is packaging it all up as timely advice that people can act on.

To help address this challenge, I recently joined EnergyLab and GrowthMentor as a mentor. EnergyLab is Australia’s largest climate tech startup accelerator, while GrowthMentor is an international platform for mentorship around startup growth. Both are relevant to my focus on helping leaders of climate/nature tech startups ship data-intensive solutions (including AI/ML, data science, and advanced analytics – this stuff goes by many confusing and hyped-up names).

The rest of this post presents some of my reflections on packaging advice and expertise. I’m always happy to discuss these topics and connect directly with people I may be able to help, so please feel free to reach out with feedback.

Actionable and timely advice by example

We all know we should get enough sleep. But telling a busy insomniac with young children to “sleep more” isn’t actionable. It’s more helpful to provide them with specific strategies for improving their sleep hygiene, like keeping screens out of their bed. And the more specific, the better: “At 10pm, put your phone to charge in another room” gives them exactly one thing they can do tonight.

There is more to it, though. If your insomniac friend comes to you complaining about having a bad night, they’re probably not expecting advice on where to charge their phone – at least not in that specific moment. The timing of the advice can make all the difference between them following it and them doing nothing (or worse – getting annoyed by your lack of empathy).

The same goes for advising anyone about anything.

In my case, advice of a general nature like “your data should be clean, relevant, and plentiful” is nice – but it’s also kinda useless. Getting more specific on strategies and tools is better, e.g, “consider dbt to manage and test data transformations”. Getting to the root of what they want to achieve may yield completely different advice, like “don’t worry about dbt for now, if you want that ML project you mentioned to succeed, you need to instrument and start collecting data around feature X of your product as soon as possible.”

On listening and packaging expertise

To get to a point of giving timely, actionable advice, you need more than functional expertise. It’s important to listen to what the other person is saying (and not saying), and figuring out what they’re most likely to respond to. This is easier with people with whom you’ve already built a relationship than with new acquaintances – which makes the challenge of mentoring at scale all the more interesting.

One key aspect is aligning on expectations. Coming at it from the mentor side, I aim to be transparent about where I can and cannot help, so as to only attract mentees who are likely to be a good fit. However, after almost twenty years in the tech industry and over a decade in data / AI / engineering roles with startups & scaleups, it’s hard to succinctly describe my area of expertise. For example, I liked the label data scientist when it became popular around 2012, but both the label and I have changed over the years. There are major differences between my experience and that of a new data scientist who is fresh out of university. Me using a commodity label like data scientist is not in anyone’s best interest.

Aligning on expectations is easier in close long-term relationships. In our professional lives, such relationships are commonly formed when working for one employer at a time. Indeed, most of my work experience was that of an employee. And like many employees with long-term roles, it was easy for me to identify opportunities for improvement and provide actionable advice to my colleagues. There is a lot of implicit listening going on when you are dedicated to a single employer!

In the absence of a long-term relationship, it’s important to communicate expectations ahead of time. For example, this is what I put in as my “support offered” for EnergyLab founders:

Advice on data strategy, data hiring, AI/ML projects, data science, advanced analytics, and data-intensive solutions.

I have over a decade of experience in data / AI / engineering roles with Australian startups (most famous: Car Next Door / Uber Carshare & Orkestra), international scaleups (Automattic /, and big tech (Intel / Qualcomm / Google). This means I also have many opinions on tech and startups beyond my specific expertise, which may be of use to some founders. :)

In a case of a potential fit, the next step on my end is to listen. My aim is to only offer mentorship in situations where I add value. Redirecting founders to others in my network who may be a better fit than me is a better outcome than attempting to give advice on topics that fall outside my area of expertise.

True experts are always learning

Another key aspect of providing advice as a mentor/expert is recognising that no one knows everything. Even within narrow areas of Data & AI, things are moving so fast that even the most knowledgeable people have no chance of keeping up.

However, expertise is a relative term. I know more about shipping data-intensive solutions than a non-technical CEO, so I can probably help them (especially if they don’t have in-house data experts). I know less about PyTorch internals than an ML engineer who has been focused solely on deep learning for the past decade, so I’ll defer to such experts when deep PyTorch expertise is needed.

As another analogy, consider a general practice doctor named Amy – she is a medical expert in comparison to most of the population. But Amy wouldn’t try to perform brain surgery – she’ll refer you to a neurosurgeon (Barbara?), who is an expert in comparison to Amy.

Things are fuzzier in the unregulated software and data worlds. Memorably, the young child of a past manager one day announced: “My computer has data on it! I am a data scientist!” The equivalent of such pronouncements in the adult world was the swift shift of LinkedIn titles in the years after 2012 – peak data science hype. By contrast, declaring yourself a medical doctor will land you in prison in many countries.

In the absence of regulated data expertise (which is probably undesirable), we are left with heuristics for determining who should be providing data advice. One of my favourite heuristics aligns with GrowthMentor’s core value of humility. In their words: “Nobody knows everything and we should all be open to hearing a different perspective on what we are working on. […] Opening yourself up to feedback from your peers will not only make you a stronger person, but also lead to more confidence in your professional life.”

To me, this is the sign of a true expert: Knowing that you still have a lot to learn. And this brings me back to what I’m aiming to learn and improve through mentorship: Giving timely, actionable advice outside the context of employee-employer relationships.

I’ll report back on how it goes in the future.


    Public comments are closed, but I love hearing from readers. Feel free to contact me with your thoughts.