Four Ways Conversational AI Companies Are Approaching Platform Access
Does Certain Approaches Work Better Than Others?
Introduction
Obviously there are a few ways organisations are approaching Language Technology tooling, and different verticals are focused on.
In broad strokes, these categories are:
- Open Source frameworks (Rasa, MindMeld, Botpress, etc.)
- SaaS/Cloud end-to-end bot development frameworks which facilitates voicebot enablement. (Kore AI, Cognigy, IBM Watson Assistant, Oracle Digital Assistant, One Reach AI, Boost AI, etc.)
- Niche Players. The platforms focussing on the seven vertical vectors of Conversational AI. (HumanFirst, Botium, QBox, Voiceflow, prodigy, etc.)
- LLMs. Lastly, Large Language Models. (co:here, AI21labs, OpenAI, HuggingFace)
How are these platforms managing and allowing access?
Platform Access Overview
In Conversational AI, platforms take different approaches to granting access to their functionality and interfaces. There are 4 broad approaches.
Full & Open Access
Cognigy took the lead in this department with granting full and unlimited access to their framework. No credit card required, no commitment, just share your email address. This move is bold, it creates great exposure, and it establishes a base for good marketing. Kore AI followed by also giving free access. Kore AI’s documentation is also very astute and a great reference for Language Technology in general.
The open source platforms are obviously part of this category like Rasa, MindMeld, Botpress, etc.
Access With Onboarding
There are a few organisations which do not have open and unrestricted access to their platform, but rather opt for an onboarding process. In a sense this is a generous approach in giving customers time, understanding their needs and walking them through the product.
Based on the engagement, next steps are useally determined. Sometimes a NDA is required, specific use-case access is granted, or specific functionality for a period of time, etc. HumanFirst follows an onboarding/demo approach which can be booked via their website. QBox grants access to a scaled down sandbox environment, but a demo/walk-through can be requested.
Access With Credit Card Registration
Enterprise Cloud Solutions in most instances demand the registration of a credit card. IBM Watson Assistant, AWS, Oracle Digital Assistant and others. Care needs to be taken when prototyping not to run up a bill, and most AI functionality does not find itself in the free tiers.
No Access Consultancy Based
This seems to be the approach followed by Boost AI, OneReach AI and others. Sometimes a NDA is required, which I personally find as an impediment, as it is not possible to sign a NDA with multiple vendors.
Organisations can have functionality or tooling which is not customer facing, or which does not have a user interface. This could be functionality which is not wrapped in a no-code UX. In of itself this is not a problem or any detraction of the product, and these vendors do often find themselves in a more corporate, consultancy space.
Playgrounds
Rasa is the only chatbot development framework with a playground. Large Language Models (LLM)bears a burden of being seen as inaccessible and highly technical. The idea of Playgrounds break this perception and grants easy access.
All the LLLM providers have a playground which is freely available on email registration. OpenAI and co:here allow for fine-tuning without charge.
AI21labs has very good examples and implementation scenarios on how LLMs can be used within an everyday production environment.
Conclusion
Obviously each vendor and technology provider will follow a strategy which suits their product and business objectives best. And each vendor will seek out a competitive advantage. Those leading in table stakes might have a propensity to be more open. Those fearing abuse, especially with the larger cloud providers will place checks, balances and some form of accountability in place.
Be these as it may, access to technology, services and knowledge is in abundance and ever increasing. With most vendors, if you reach out, and are willing to learn, knowledge and insights are usually freely shared.
