Where does the Chatbot sit?
Geoffrey Moore
Author, speaker, advisor, best known for Crossing the Chasm, Zone to Win and The Infinite Staircase. Board Member of nLight, WorkFusion, and Phaidra. Chairman Emeritus Chasm Group & Chasm Institute.
As the platform software players continue to expand their footprints—and their market caps—they are turning more and more to AI and ML enabled services to extend their reach and effectiveness. The TAM (Total Addressable Market) here is near infinite. Two centuries of dedicated development of the industrial business model, where product is king and the customer has to stand in line, has created a ubiquitous landscape of inefficient customer-facing processes. Or, as Milton once wrote, “They also serve who only stand and wait.”
Well, we can all see that’s changing in a big way. Like ATMs or any other self-service transaction, chatbots, and their various cousins, have the power to provide better service, faster, more reliably, and without social negotiation. There is a downside in the elimination of a current generation of jobs, but there is reason to believe that all we are doing is moving the frontier of human-centric creative services to a new boundary. No one believes that there is a future in which everything just works. The challenge is figuring out where the next breakpoint is and skilling up to meet the emerging needs.
That all said, there is an immediate issue facing the SaaS provider today. How do you charge for AI/ML services? The prevailing SaaS pricing model has been a per-seat subscription fee, but chatbots don’t need seats. Moreover, they eliminate the work of people who do. The result is a self-imposed conflict of interest that must not be ignored. We have already seen this kind of pricing problem play out to negative effects in the business process outsourcing sector, where companies delay the roll-out of automated solutions, or introduce unnecessary complexity, in order to keep their current staff billable. We don’t want a repeat performance in SaaS.
The good news is, it is relatively straightforward to address mismatches between value delivery and pricing models—you just adjust the pricing model. The bad news is, a whole lot of forecasted revenue and earnings have been predicated on extrapolating from the prior model, not to mention a boatload of comp plans for the sales force. In the short term there is likely to be a shortfall. That causes a pause, in effect, an invitation to a next-gen disrupter to attack your market. And rest assured, when they do, they will introduce a pricing model that exploits your vulnerabilities to the hilt.
In short, this is the sort of challenge you want to get the jump on early. Yes, there will be some sacrifices to make, not to mention a bunch of change management to work through. But the longer you wait, the more constrained you are. And if you wait too long, you are well and truly trapped.
That’s what I think. What do you think?
Follow Geoff on LinkedIn | Geoffrey Moore Mailing List
__________________________________________________________________________
Geoffrey Moore | Zone to Win | Geoffrey Moore Twitter | Geoffrey Moore YouTube
Board Member, Chief Information Officer
3 年Geoffrey Moore great insights! I truly believe the SaaS industry will shift more to value delivery which manifest more in the form of usage based pricing model across the industry. We never pay for how many electronic device we connect to the grid at home, but for how much power I use to get value out of those devices. This paradigm will become the new norm for SaaS!
Head of Conversational AI @ Zurich Insurance | Generative AI | RAG | Product Manager
4 年How do you charge for AI/ML services? Good question. Here's my take. AI/ML can't be charged out of the box because most good AI/ML solutions need to be tailored to the customer to provide value for the customer. The few cases where you don't need customization will be commoditized by Google, Amazon and Microsoft if they have not been already (the low-hanging fruits in image recognition, NLU intent detection, speech detection have been done). So to have a fighting chance against the larger players an AI/ML provider needs to have a product that automates as much as possible and leaves a small piece for integration / tailor-made service. At AlphaBlues?we've built numerous deeply integrated chatbots for the enterprise and see this a roughly 80-20 split. That 20 is the fine-tuning that makes the AI/ML specific to the customer needs i.e. icing on the cake. For now we see 2 ways of pricing chatbots - on volume or complexity. Volume is an easy metric but tough for the customer to comprehend and predict (price may look small in the beginning if usage is low but if usage grows the price might get too high). Complexity is a better measure - predictable for the customer with no surprises, usually dependent on the capabilities of the chatbot (e.g. can it authenticate users, provide personalized info via API, on-premises hosting).? In terms of live chat then per-seat pricing will not go away because chatbots will not automate all of the customer service agents' tasks. There will be human agents and per-seat pricing in live chat. What you might see is a 20% reduction in the number of seats required (as chatbots fully automate that from natural language queries) but tough and detailed questions will require a human agent.? Aside from cost-cutting (reduction of labor costs), sales chatbots are where AI/ML pricing can have a more significant impact. Providing personalized offers at scale for users (a la "thanks-for-chatting-by-the-way-here-is-the-new-iphone-you-can-upgrade-to-from-your-current-iphoneX") will bring in revenue for companies that previously was not there. So capturing a slice of that as a % of sales is where chatbot pricing can turn really lucrative.
Head of Sales Enablement and Productivity, Americas and Global & Startups. Head of Global Onboarding | ex-AWS | ex-Microsoft
4 年There are so many areas that require a human being that we have never had enough time or appreciation for. IMHO automation and robotization of functions where you don't need to use the full capacity of your brain is a good thing. That means we will have more time to invent and explore.??
Interest PoV. One could argue that personalised services (access to your account data) requires a ‘named seat’, albeit at a far cheaper price point. The organization I work at decided to give away chatbot access as a value add to existing per seat licensing which makes sense if it is a supplemental channel.
It's a rare opportunity when I'm between jobs! Please get in touch with me if you are hiring, or know someone who is or will be. Not in a hurry, I'm just seeking the right team and offer.
4 年Let me know if you need research on chatbots (virtual agents) or even toy robots - we cover all emerging technologies! #AI?#robots?#robotics?