ThoughtSpot vs PowerBI copilot
I came across what seems like a ThoughtSpot PR video comparing it to PowerBI copilot, and I take the opportunity to comment on both of them
I came across this video comparing ThoughtSpot and PowerBI copilot on my LinkedIn feed. And I thought this gives a good opportunity to talk about both in one blogpost.
Now, this video came to my attention thanks to someone from ThoughtSpot who I’m connected to, and so you know what it is going to say. Nevertheless, it gives some very useful info in terms of both technologies, which I can compare to our (hypothetical) product.
(Looks like it’s not trivial to embed a Loom video here, so just use this link if you want to see the video. It’s a 5-minute long screenshare).
Anyway, some pertinent observations on both platforms:
PowerBI copilot is not “on premise”. I guess there is nothing new in this, given it uses the OpenAI API. The thing I learnt from this video is that in most regions, the copilot processes data outside the customer’s geographic region
PowerBI copilot is a retrofit of generative AI on an existing platform, and it clearly shows in the documentation (shown in the video). If you think about it, PowerBI is all about creating static charts that can be part of your static dashboard. And the copilot allows you to create these static charts using natural language prompts.
Interesting thing is it doesn’t let you modify the charts using the copilot. I assume you’ll be able to do that with the “normal” drag and drop PowerBI features?ThoughtSpot is the OG “analytics copilot”. They were doing Q&A on your database using natural language well before LLMs became a thing. So if you think about all the zillions of “copilots” being built, they are all basically using GenAI to do what ThoughtSpot has been doing / claiming to do for a decade.
Very quickly the limitations of the copilot model come through. The person making the video says “show me top sales”, and it shows the sum of sales. The thing with the copilot model is it is “several shot”. You need to have the patience to interact with it over multiple iterations before it gets what you need (which makes it bizarre that PowerBI CoPilot doesn’t offer this functionality to change things. Maybe some “trust and safety” issue?).
I feel MORE vindicated on not choosing the copilot routeThoughtSpot is UGLY. Check this graph that I screenshotted from the video. I know it’s in response to the user’s query on the cuts he wanted, but coming from a ~10 year old product, this kind of UX is crazy (obviously not in a good way).
Despite “BI” being in my title in my last job, I’ve never really used BI tools, but whenever I’ve come across them my instinct has been that they are ugly (my reference being MS Excel). Tableau is supposed to follow the grammar of graphics (check out my podcast episode with S Anand where we talk about this), but it’s still ugly.
And ThoughtSpot is no better.That said, for a pre-LLM product, ThoughtSpot does a fairly good job of parsing queries and understanding them. Then again, given that this has been their bread and butter, I wouldn’t expect anything else.
Over the weekend, I was at yet another networking event. My elevator pitch actually got worse there, as it was a fairly technical audience (this was a meetup for entrepreneurs from IIT Madras) and they kept finding holes in what I was saying.
I got a lot of “why man, this is such a crowded space”. And since then I’ve been thinking of whether I’m actually operating in a crowded space.
And the more I think of it, I realise it’s a matter of positioning and branding. The space that is truly crowded is that of the analyst copilots - you have the OG ThoughtSpot, and then you have a whole bunch of LLM-driven agents which tell you about your data in a Q&A format (with all the downsides that come with a copilot).
With our anti-copilot positioning, we are clearly not there. And I don’t know of too many others who are going with an anti-copilot product in this case. The problem, I realise, is that we are not communicating this well enough (cue my travails with my elevator pitch), which is leading to people slotting us in the “AI for BI” bucket, which they understand as “copilot”, and then dismissing us as “operating in a crowded space”.
Now I’m regretting my C in Marketing 101.