There's no denying the allure of the dazzling new AI products entering our industries and expanding what we thought was possible. But as most of us are finding, the promise versus reality differ greatly. Anyone experimenting with AI-led tools will become familiar with the effort of repeatedly checking, correcting and checking again, to the point of AI fatigue. How many have found their ChatGPT usage plummeting (as mine has, off a cliff)?
This brings to the forefront the question of how Generative AI will be embedded in our daily operations in the future. Will our processes begin and end with an AI chatbot like ChatGPT? Or will it be more subtly integrated into our existing products, so we barely notice?
My thoughts centre around the argument for Horizontal SaaS versus Vertical SaaS - a concept rarely discussed in regulated verticals like wealth management, where compliance costs are high yet market size is smaller than industries like healthcare.
To explain, a horizontal SaaS product is highly specialised at specific tasks and aimed across industries. For example, a CRM like HubSpot, mailing tools like Mailchimp and a website builder like Webflow. These products are usually exceptional at their particular functions, requiring hefty investment to maintain product development and marketing leadership.
In contrast, software focusing on a niche task in a vertical segment like wealth management, legal services or accounting practices such as reporting, features tailored, templated tools to assist users. Vertical SaaS is tricky to perfect for several reasons. It necessitates specialist subject expertise, either through user collaboration or knowledgeable professionals. It usually needs anchor clients to help fund the software. Significant development and features are required before profitable pricing. Given smaller market size, high supplier competition poses risks. However without this innovation - the industry stagnates.
So what does this mean for AI? Substantial risks. Generative AI relies on quality data to produce meaningful results. Many new GenAI products are horizontal SaaS - tools created for a specific purpose using LLMs trained on generic data. For heavily regulated industries, tools require meticulous training and customisation- something better suited to a vertical SaaS solution.
For GenAI applications - we recommend those in a niche vertical carefully consider their use case and the supporting data schema. Be cautious of the request and understand true benefits. Is it AI the user wants or is the requirement an automated workflow? When testing AI-driven tools, ensure they are sector-specific, trained on relevant data to avoid industry pitfalls and language specifics that only a tailored tool would recognise.
Looking at the broader GenAI prospect for Wealth Management - existing Vertical SaaS solutions with the most data are best positioned to enhance technology, yet also least likely to do so. Quality user data is key before entering the AI era. Most vertical SaaS incumbents have significant work ahead on client data integrity before integrating tools that will expose their adoption challenges. This presents a once-in-a-generation chance for new innovators with the promise of efficiency through emerging technologies. These innovators also face risks in finding early adopters to achieve mainstream success. If ever there was a time, it is now, but is the technology up to the task yet? And how will the industry overcome its biggest efficiency challenge - poor data quality - that persists despite new technology?
My wager is on Vertical SaaS with embedded AI. How to get there? Twelve months of diligent effort on data cleansing, enabling the next generation of maturing, learning technology to advance.
In publishing this, I had a great opportunity to test the discussion in this article. LinkedIn offered to design the picture for the blog with its embedded Microsoft AI tool. With Embedded AI, my journey was slick with no need for integrations or changing tools. Unfortunately, it misunderstood the title of the article, so I had to do the design myself in Canva and manually load it back...