Throxy - Sales is changing fast.
A prospecting assistant powered by AI - sales teams shouldn't be writing emails anymore.
This week is focussed on a company called Throxy. As we have all read a million times AI is going to seismically shift the way we all work - it’s going to eat a lot of jobs and create a lot of new ones, jobs aside, productivity is going to increase exponentially. This paradigm shift should be welcomed, Throxy is a great example as to why. Before we get onto talking specifically about Throxy and why it is such an interesting idea I want to outline why I am feeling so excited and generally positive about the AI paradigm shift.
This comes back to a core belief of mine that most people aren’t dick heads and that yes just as the internet and general technological improvement did, AI will create bandwidth for evil actors to potentially do more “harm” but it will also allow the majority of people (who as stated previously, in my juvenile, poorly informed, western centred opinion, generally aren’t dick heads) to create incredible positive change and general improvement. I understand the comparison between the dawn of the internet and AI isn’t bulletproof but it is worthwhile because they are both tectonic shifts that aren’t confined to a certain geography, industry or type of people, they both have and will touch everyone and everything. To sum up - I believe AI will be net positive, I am glad the inevitable has already happened; the genie is out the bottle. We have begun the great gamble, can we harness AI and improve everything about our daily lives without letting it spiral out of control? I believe we can because of the very simple fact that I think there are many many more good people than bad ones.
What is Throxy?
“Throxy is a sales outreach platform powered by hyper intelligent AI agents for research, copywriting and smart scheduling.” (Pablo, Throxy CEO & Co-Founder)
Their primary focus at the moment is to train AI agents (large language models) to automate/improve/personalise email outreach at scale all while ensuring their clients company domain remains in tact.
Each Throxy customer will have their own custom AI agent that truly understands the type of prospect that the Throxy customer is looking to reach and the pain points Throxy’s customer solves for their end customers. This tuning is absolutely vital to getting max value from the product.
The problem they are trying to solve?
Eliminate wasted time spend on the sales team manually writing prospecting emails.
Retain/improve the health of the company domain to ensure emails don’t end up in spam - regardless of how many emails are being sent out.
Then the golden goose - efficiently train AI models/agents to ensure that each email sent is highly personalised to the company and the person receiving it to improve reply rates and exponentially increase the amount of qualified initial meetings you book.
“Throxy can send 10 thousand emails in one day to your prospective clients and each email would be totally personalised to the recipient and in my opinion much more effective than the 15 prospecting emails you typed out manually, it’s that simple.” (Pablo, Throxy CEO & Co Founder)
Throxy Co-Founders:
Pablo - Throxy CEO & Co-Founder
A biologist by training, turned salesman. Having worked at Kanjo, Unibuddy and Showpad as a business development representative, Pablo experienced the issues first hand. Sales teams either manually send very few emails or they hire a third party that sends huge amounts of emails that ended up either in spam or not being read because they were so obviously broadcast emails that had been written by a bot to be sent to hundreds/thousands of inboxes.
As I have said before on this Substack the best problem solvers are people that have been in the trenches and experienced that problem viscerally. From my few calls with Pablo, in my opinion he is a blend of magnetic in his drive and passion (a natural salesperson) but also clearly very intelligent and therefore truly understands the issue/space Throxy is entering.
Arnau - Throxy CTO & Co Founder
A computer scientist and AI engineer. He’s been working with LLMs since BERT and GPT2 and has extensive experience fine tuning them from when he was at JP Morgan.
I asked Arnau why he took such a big risk and left JP Morgan - this was his answer:
“I've been close friends with Pablo since we met in school back in year 7. I was always tinkering and hacking projects on the side, I was an early adopter of LLMs, working with GPT-2 and other early versions at JP Morgan, Pablo knew this and wanted a tool to help him to deliver more value in his SDR role, so he asked me to see what I could do. So I began building him the tool that was basically an early version of what Throxy is today, I loved doing it. All that aside, I believe that in a world that is changing so fast (especially now with AI) not taking any risks is the biggest risk you can take so we had to jump all in on this.”
Domain expertise + technical expertise is the formula to solve real life problems and create a huge amount of value. At pre seed - the founders are the name of the game - idea’s change, markets move and product/services pivot - if you hand cash over at this stage you are betting on the people founding the company and these two are good pony’s to be backing.
Fundraising:
They raised a pre-seed round of £200k in December 2023 from Venrex Investment Management - that money is being used to build a small technical team and improve the product.
One of the great things about having a product like Throxy is the tool can obviously be used internally - currently they are growing very quickly using their personalised email outreach and word of mouth - all with a minimal sales/marketing spend.
There are no immediate plans on the horizon to raise a seed round, however at the rate they are growing this seems inevitable.
Market size, competitors and incumbents:
I will keep this fairly brief, Throxy can work with most/if not all businesses with a sales function, so the market isn’t small!
When it comes to direct competitors they are tough to find as AI and automation/personalisation at this level is still such a young part of the industry. My research has led me to two real competitors that are both slightly bigger than Throxy and in some ways focus at different parts of the sales process, they are called Glyphic and Clay. Both look like impressive outfits and I look forward to seeing how each progresses.
With regard to incumbents Pablo had this to say, “Most of the incumbents are sub 10m ARR - they take data, put it through chat GPT and unsuccessfully produce “personalised content”
What’s next for Throxy and what are their biggest challenges going forward?
I think the focus over the next 6 months is to find product market fit with their personalised email outreach, data management and general prospecting automation. To then start driving towards becoming an end-to-end prospecting platform. AI agents to find leads, AI agents to research them, AI agents to copywrite the perfect message and fully manage the sequence of emails sent. They want to automate all of this so that the human is only involved once the prospect actually wants to meet either over a call or in person. So find product market fit with one vertical and then scale out the ecosystem to ensure they solve for the problem in a more holistic way.
Pablo highlighted that it’s a “tough pill to swallow” for sales people, the thought of a machine being able to push out thousands of emails that are better written and more personalised than the 10 emails they manage in the couple of hours per day they assign to prospecting emails. Once people experience the power of the product the customer becomes sticky, however it is a challenge to get teams to trust that the automated, AI driven route is the way forward in the first place.
Arnau outlined that on the technical side of the business the main challenge is firstly to better understand the most efficient way of building in this space. LLM’s are still so new that there are very few well trodden paths, the team are constantly having to innovate and find their own way. “You can run 1000 tests in your staging environment and then still get an unexpected output in production with LLMs” (Arnau, CTO) The problem they are trying to tackle going forward is to reduce any sort of human input but retain the high standard of output, “we want the agents to have more and more autonomy”. (Arnau, CTO).
Conclusions:
Throxy is a great idea/technology, with a massive addressable market and two impressive/driven co-founders. I haven’t gone into much detail on their competitors as this post is already too long but from what I can see Throxy is up against some A team players building in the space. That said, I don’t think that’s surprising at all as this wave of companies represents the next big step in sales tool technology, there was always going to be stiff competition, just as there was with the CRM boom.
To finish I thought I would highlight a couple of interesting anecdotes from my interactions with Arnau before writing this piece:
My question: When the level of compute available and the cost of compute inevitably goes down what part of Throxy's offering will improve and how? The processing of the data before it is put into the model? The vectorising? The accuracy/reduction in hallucinations?
“Higher levels of enrichment, ultimately with compute costs going down we will be able to run faster and better models which will lead to having specialised highly niche agents doing incredibly specific tasks. This will improve accuracy and reduce latency. Regarding vectorising if the promises around long (potentially infinite) context windows work out and compute scarcity continue to decrease I think we will be shifting away from vectorising more and more” (Arnau Throxy CTO & Co-Founder)
To give some context, they vectorise data sets before using the data for their models to ensure they are using tokens efficiently and reducing the amount of hallucinations the model produces.
My Question: Are you more excited or concerned about the rate at which AI in general is improving?
“Super excited it's the best time to be alive, we are accelerating society as a whole and I am confident (hopefully in our lifetime) we can see AGI and many positives coming through. We are relocating human capital to work on greater and greater problems.”
These two are classic examples of people that want to build in the AI space to produce net good and create a lot of value! My kind of people!