Platform Agnostic Conversational AI
AI is defined as, smart machines that extend human capabilities by sensing, comprehending, acting and with continuous learning and thus allowing people to achieve much more than what a normal (or shall I say an intelligent) human being can achieve!!
Let us think of set of technologies that are encompassed by AI and some are as given below:
- Virtual Agents i.e. Human Machine Interface, helps in establishing first tier of contact with customers, responds to queries and questions with non-verbal cues and replaces online customer service representation
- Robotic Process Automation, (I know some of you might not agree with me putting RPA under RPA)which put in simpler terms is like mimicking human actions (transactions). Imitates routine, time consuming tasks, Viable for high volume processes, requires rule-based decisions without judgements and provides digital triggers and inputs.
- Natural Language Processing or speech/text comprehension that enables understanding between human & machine, comprehends nuance of language(s) and extracts mean and expresses messages in natural language
- Image recognition that helps in ocular analysis/understanding. Identification and detection of objects or features from digital images and requires large learning data set
- Machine Learning that helps in big data pattern recognition which involves aanalyzing large data sets, aligns with subjective, malleable processes & expectations and predictive & prescriptive and interprets text, speech, visual, and digital inputs
Ok, now you must be able to relate some of the above with the kind of products that you see in your daily life, for example AI enabled IVR, Home Automation, Alexa, Google Home etc.
I have realized one important aspect of the whole Conversational AI product landscape after going through so many technical articles, you just can’t rely or be over dependent on one set of products and rather have a very platform agnostic approach if you would like to be highly successful in this space (and you know what, it is relevant for most of the technologies these days).
I will try to cover some of the aspect of how this space is staying away from product dependencies and why is this the best approach.
The conversation AI product ecosystem is rapidly evolving, the trick is to keep adding more partners in your ecosystem to refine/re-define your platform agnostic approach. Do evaluate them before teaming with them as it is critical that you have right set of partners to go to market with. Some of the specialist and leaders outlined below.
What are the key capabilities that you should look out for in the platform?
It should be channel agnostic; customers can access any/all channels of their choice and this platform can integrate with these channels. How do you achieve that? The microservices Channel Adaptor can quickly integrate to your platform and connect to your most common customer-facing channels, enabling Virtual Agents to reach customers across a range of interaction points.
NLP-training functionality in the Engineer Portal, that will allow you to view, re-classify and commit training data directly from your platform to your selected AI Provider’s NLP models in the cloud or on premise. Examples are IBM Watson, AWS, Microsoft, Google DF, RASA etc.
Intelligent agent console, to empower the human agent and improve customer satisfaction. Complex or escalated conversations can be seamlessly handed over from bot to human.
- Features include an Agent Coach, with query classification and conversation tips, as well as Customer Profile (with personalised offers) and Local Store info.
- Through a collaboration with Search Technologies, accessed through the intelligent Agent Console, the Knowledge Search and Curation capability enhances the quality of the human-agent’s responses.
- The agent can get a 'Quick Answer' and 'Useful Notes' for follow up questions, highlighting the exact information required to solve a query.
The Developer Portal enables Bot configuration, management and reporting though a friendly UI. Developers can switch (and train) their NLP model, view conversation logs and add new use-cases.
The Business Operation Portal displays key performance metrics, including Automation Rate (and associated financial saving / FTE equivalent), Customer Satisfaction, Average Handling Time and Customer Time Saved.
Just knowing the key capabilites won't take you anywhere, as I said in the begining you would probably be looking for a partner that can serve you well keeping in mind the platform agnosticism is the demand of the hour!! Let us also check on the key attributes of any successful or reliant Conversational AI service provider.
- We are surely looking for a platform that require “Low to “No” coding. You can have Bots built by the decision scientist/ business analysts without writing any code.
- Perpetual license provides access to source code, allowing you to enhance & customize the platform as you like, without lock-in.
- Multiple bots for different domains or different geographies can be hosted in single instance with bot level access to business authors, thus enabling “Multi-Tenancy”.
- Access to domain specific “data libraries” have been developed to provide working use cases for common flows.
- Different commercial & support models available to support different client needs and domains/geographies.
- The platform should provide production ready modules for bot design, execution, analytics & agent escalation for end to end lifecycle of bot interaction journey.
- Ability to use best in class AI services from different vendors (cloud / on-prem) to an integrated bot solution, switch between vendors when the need arises.
Just today, while in one of the sales call, the prospects was not happy with the platform that was implemented at his site by a well-known name in this space and he was not at all happy with the ROI and the time it is taking him to realize the benefits, his main pain point was the key business KPI’s are still not being met and the reduction in resources and ultimately the cost is still not realized. So the key question would be “how would one measures the success of its Conversational AI?”.
A variety of KPIs can be measured in pilot or production to evaluate the success of a Conversational AI solution.
- Average handling Time = this is reduced by AI (reduction in queues, faster responses), ultimately leading to live agent time saved, leading the employee satisfaction and better overall business ROI.
- First Contact Resolution = resolving customer’s problem 1st time – in %, resulting into customers time saved and thus customer satisfaction and better NPS.
- Deflection / Automation = in number/ and $ saving, helping you cover more queries and address more customers, reducing the overall wait time and again improving the CSAT, NPS, ESAT and helping with better business KPI's.
- Customer Satisfaction (that is the results of reduced AHT, Queries less deflected, FCR)
- Employee Satisfaction and many more..
Demand has evolved in the last few years which conversational AI trending on combinatorial opportunities in the space of knowledge and management curation, IVR, voice biometrics, IOT integration, contact centre transformation, Blockchain, so keep looking out for disruptive changes in this space with Video KYC and Video Analytics also coming into this overall product landscape.
coFounder @Exotel | Driving Growth at Exotel | Connected Customer Conversations #LikeAFriend
4 年Well written Sumit K.. Conversational AI has shown good promise and I think the carefully done experiments are now showing results in the form of KPIs mentioned. This has come a long way from a fidget spinner in the hands of the CTO to a strategic direction for a Strategist.