Era of the Bots

In enterprise conversations of 2020 inevitably one would come across bots in various contexts. References to Chat Bots, RPA bots, CPA bots and next gen Multi Functional Cognitive Agents will be sprinkled across conversations involving enterprise productivity, efficiency, automation and user experience. What are these bots?, How are they different? what does the future hold for the bots as well as the enterprise?

A bot is typically a software application to do a specific task, often simple and repetitive. If you consider level of sophistication or intelligence as a benchmark of progress, the bots can vary from rudimentary desktop automation bots to highly sophisticated multi functional cognitive agents. With chatbots being somewhere in the middle end of the spectrum as a transitory technology.

The image below shows some of the common technologies and automation terminologies in chronological order

RDA Bots (Robotic Desktop automation)- Rudimentary screen scraping technologies, created as a bridge between current systems and incompatible legacy systems. Takes over the bulk of repetitive data entry or form-filling tasks from humans.

E.g A standalone desktop application or a client server/thin client system without any APIs or eternal connectors available that require a human agent to go through form fields on a screen. Typically use the XY coordinates on the desktop screen to retrieve the data and enter into another system (read newer technology with APIs available) – works with manual intervention. Prevalent technology since the 90s.

RPA Bots (Robotic Process Automation) – Desktop automation with added digital triggers or self-service (unattended), Deals with only structured data, predefined activity choreography to do the tasks repeatedly. Started becoming popular since early 2000s

RDA and RPA function at a repetitive task level and value adds the enterprise in reducing the workload on the employees.


Chatbots are usually scripted decision trees where a set up tasks are completed in a top down flow with button clicks and filler sentences. Chatbot handles simple repetitive chats where the chatbot creator scripts user chat and bot responses.

CPA (Cognitive Process Automation)

Using AI to replace human intelligence/cognition related work. Works with unstructured data along with natural language understanding and generation. Producing insights, analytics, and taking actions at or above human capability. Started Mid 2019

A RPA Bot works by recording the user’s action on the screen and repeats the same set of actions. I.e mimics human repetitive actions while Cognitive process Automation mimics the Human cognition. For simplicity, one can think of RPA as a software robot(bot) that mimics human actions, whereas CPA is concerned with the simulation of human intelligence by machines. Cognition refers to gaining knowledge and comprehension – cognitive processes include knowing, remembering, solving, etc encompassing language and perception.

Cognitive process Automation automates processes that require human cognition and the process of attaining this cognition is AI training. At a very fundamental level RPA is associated with doing and CPA is associated with thinking and learning,

Chatbots are transitory technology, since chat itself is a cognitive function, the current set of chat bots with scripted inputs and outputs will not be scalable to the next level of cognition due to the limitations of the intent classification systems deployed and the limited learnings on the user actions. Chatbots were relevant as a stop gap solution while complex cognition AI models were getting matured and giving rise to Multifunctional Cognitive Agents which are capable of Cognition with unstructured data (Voice / Text / Images / Video/PDFs/Xlsx/..), Structured data and Free Flowing conversations

Multifunctional Cognitive Agents (NLPBots) are capable of functioning as fully cognitive business assistants at or above human capability.

I.e for the first time in human history there is a competition to humans from Multi functional cognitive agents for a place in the enterprise workforce. Multifunctional Cognitive Agents – NLPBots uses various cognitive capabilities (reading Unstructured text, structured data, scanned documents and pictures, videos etc.)  to perform various tasks as part of the enterprise workforce. Just like training a new employee one can add various data sources to NLPBots platform and create Multifunctional Cognitive Agents – As an example, NLPBots inducted into the enterprise as a receptionist recognizing and welcoming people, checking their documentation, booking appointments, etc. (It involves face recognition and identification, document scanning and data extraction and validation, conversations to complete a task, etc.. ),  or NLPBots could be trained to generate reports like an analyst from available data. Or It could be for welcoming customers and solving their problems. It could be an NLPBot in AIOps, ITSM, Employee Engagement, or something else altogether. When there are various cognitive capabilities like reading Unstructured text, scanned documents, pictures and videos, coupled with the ability to train a limited amount of data on the platform to create new AI Models creates immense opportunities for Multifunctional cognitive agents in the enterprise workforce of today and tomorrow.

Is your Call Center optimised for Peak? It can now

No traditional call center can be optimized for peak. But that’s what call centers should be doing in the first place.

To put that in perspective let’s take a few situations

Situation 1: You are launching a new product, you have many campaigns running across various media and a lot of information about what you want to launch is out there. It could be a product, a feature, a new service or service line -you put some information out there on that. There are ad agencies, marketing teams, on field staff, event organisers, and even your sales teams that are geared up for the occasion. They all ensure that the data you put out there is good and appealing to the market segment you are targeting. And yet… on launch, there are a few variables that you couldn’t gear up for. The response is fantastic. There are thousands of calls every minute enquiring about a feature that was there in some media but not in others, rather not clearly visible in some campaigns. Or some information that was of concern to many many but did not come up in the survey as really important. Or simply to book / order the product or service!

The call centers were trained and equipped with the details on the product but the system couldn’t handle the sheer traffic. There weren’t just enough people even though all leaves were cancelled and extra systems were put in place. The SLA capacity was half the capacity available today and yet the extra wasn’t enough to cater to the hype around the launch plus the usual regular volume of the call center. This also raises questions on the product as the first experience with the enterprises wasn’t that great for most who called in!

An excellent product took a beating on the day of the planned launch in spite of an excellent campaign, or shall I daresay because of an excellent campaign.

Situation 2: There is an unfortunate event. You had to shut certain operations for a while. Its all over the news. It was sudden and you didn’t have enough time or budget to react properly. You released a press statement but it wasn’t published everywhere.

Your current call center is jammed with calls from your existing customers. They just need someone to tell them that you will be back or that servicing will continue for a while. But they just cant get through. That gets them more frustrated and can lead to mob attacks, legal cases and more.

An externality caused irreparable damage to your brand

Situation 3: It’s a regular day. Your call center receives calls from customers / visitors on the product or service or they have a complaint or issue with. Your agents provide them with information or advise them. If the problem is not resolved they register a complaint by raising a ticket. Then the “feet on street” get involved to close the ticket.

Here basically the call center is simply managing the drawbacks in the product or service. It could be technical, functional, ergonomical, etc

Any contact center worldwide will be catering to either of the above situations any given day. These agents have limitations and so does the infrastructure to cater to situations 1 and 2 above. When actually that should be the raison d’etre for any contact center. It doesn’t actually exist to compensate the complexities in a product or service – which should be work in progress as far as the product is concerned or automation when it comes to service.

Unfortunately, the reality is that you cannot really optimise any call center for scale. The unit economics won’t just add up. Some times in cases of external entities that you outsource for the operations they may be able to add on some additional infrastructure to manage for planned events (i.e. launch, etc) for a while. But it doesn’t work out if its sudden or unexpected and the scale also is within limits. Its almost impossible to bring in 4 times the workforce, train them to answer questions and handle the software to help the customer or prospect in need.

Apart from these sheer availability challenges there is the issue of quality of the hardware, software and human resources in question.

Up until Light Information Systems launched the Cognitive process automation platform NLPBots there was hardly any alternative to the call center. The contact center was a bunch of people manning calls (VOIP or telephony switches) from terminals and some hardware or software was used to route the calls to the appropriate agent based on their availability or expertise. This system had many limiting factors including

  1. Inconsistency – depended on the mood of the agent and caller
  2. Authentication –  was linked to some pieces of information verbally furnished by the caller
  3. Information gathering inefficiencies – the caller having to repeat authentication information or complaint to multiple agents
  4. Confusing – Lengthy IVR messages which may not have the option the caller was looking for. (if we knew what the caller wanted anyways why have the call center in the first place!)
  5. Zero personalisation – its always another agent who has no clue of what the caller is going through or went through with the previous agent.
  6. Action plan – in case the caller was required to take some action there was no way to send over some documentation.

All this of course if you somehow made your way to the top of the queue and actually spoke to an agent!

Of course, since there wasn’t a viable alternative and the demand was huge enterprise, (being enterprising of course!) came up with solutions both software and hardware to compensate for the shortcomings to some extent. Cisco, Avaya, IBM, etc created these solutions to optimise the contact center. You had software that opened up once the authentication was done with some sort of history on the caller and the service, in some cases the contact center would work in conjunction with an SMS provider who could messages with information back to the caller, also Caller identification provided by the telecom service provider. But the gaps still remained and the original problem of sheer availability couldn’t be overcome. By some estimates, the losses to enterprise in terms of sales and brand ran into hundreds of millions of dollars annually.

Enter the Multifunctional Cognitive Agents or NLPBots.

Built on the cognitive process automation platform (CPAP) these agents can be trained on any data set and connected to any software systems and cater to any stakeholder external or internal and automate processes that ease their pain instantly.

Training: Even a single MfCA (Multifunctional Cognitive Agent – NLPBot) can be trained on  multiple processes and entertain multiple user groups

The NLPBots can converse in multiple languages depending on the customer’s choice.

The NLPbots are consistent and engage in the best quality conversations no matter what the mood of the caller is.

The NLPBots don’t forget. Every returning customer is greeted based on their previous experience.

The MfCAs come with automatic intent classification- which means that the customer doesn’t have to go with any IVR. They just speak / chat with the system and the NLPBot classifies the intent based on a) what the customer said b) the customer profile and c) purchase history.

The NLPBots can be trained to take action based on the conversation such as send an SMS or email to the caller with the details discussed or even the entire chat transcript. Or generate a ticket for the service teams on the Ticketing software.

And above all…


Scale Numbers: The NLPBots can be scaled anytime or automatically with the increase in usage to cater to peak. System also automatically queues requests if needed.

Scale Processes: The scale is not just in numbers of agents to manage the peak volumes of users but also in processes. Which means that if a bot is used to cater to customers by helping them with product information or service processes but can scale to engage visitors from your advertising campaign with information with the same product playbook training from customer handling. It can also “listen” to social media conversations on your product or service and take action if needed. It can even schedule appointments or even collect payments depending on the channel.

Scale Channels: The NLPbots can also scale channels and give a completely omnichannel experience to the customer. They can begin a conversation in one channel such as FB messenger and continue the conversation on Whatsapp. Of course, the customer will have to identify themselves for this seamless experience

The new age of enterprise cognition is here and all enterprises have to adapt to it. Till now the argument was the “taking away” of jobs, the lack of training data, etc. Today the COVID 19 Pandemic has alas exposed the frail nature of our human bodies and enterprises that have to survive have to be competitive and provide the services to humans with or without them doing all the work. The Multifunctional cognitive agents like NLPBots will ensure that your customers have the best personalized experience at the lowest cost to your enterprise while ensuring that they keep learning to be able to do more and more functions by adding data sets and processes. With the increase in unstructured data being generated through social media, emails, soft communication, collaborative channels, etc any enterprise that doesn’t make sense of it in real time misses out big time! Watch this space for more on that…

Cognitive process automation for Employee Centric Enterprises. Why Now?

Cognitive Process Automation – On a Platter

NLPBots is a Cognitive Process Automation platform. The world’s first and most powerful. It’s also very versatile. There are over 180 processes currently being automated using the platform. Processes that needed cognition hitherto only available among humans. When we combined NLP with AI we were able to build that cognition into enterprise systems and have created use cases ranging from AI assistants for employees to report generation for marketing insights.

Classification Of Processes by Stakeholders

Let’s just take a group of processes from the entire gamut of enterprise processes. These enterprise processes can be classified by verticals or functions. Normally that’s done because IT systems are made and sold that way for the last 50 years! Let’s take HR processes for a services enterprise. Here we will have an HCM or HRMS that keeps track of the employee, every time he / she shows up, or doesn’t, needs leave, or who he/she reports to, or makes a claim for an expense incurred for the enterprise, etc. Now if that employee is in Sales he / she will also be connected to various sales related processes like number of leads generated, volume of the sales funnel, targets etc. There will also be shared services, meeting rooms to be booked, learning and development, upskilling, etc. The CPA can automate all of these processes by removing the dependency on humans to perceive the request (made in natural language), initiate the process, retrieve information from unstructured data (Policy, SOPs, etc), pull push data into any of the systems (HCM, CRM, LMS, ERPs, etc) and post execution even give feedback to the relevant stakeholder. Since for the first time we have a platform that cuts through verticals and functions we can choose a simpler and far more effective classification of processes. We take the stakeholders. A stakeholder / user first approach has many advantages.

  1. It helps clubbing of processes based on the user that most benefits from it.
  2. It gives that user that omnichannel availability he/  she needs to make his / her life simpler.
  3. Any department can create use cases that automates their processes relieving humans in the loop that may have been non essential in the first place
  4. System can generate insights based on individual and group interactions.
  5. Even users can be connected with each other via a common platform to achieve more.
  6. The classification can be universal and will apply to every industry vertical

So broadly we classify all processes under Customer centric processes and Employee centric processes. Also there are other stakeholders like management, vendors, franchisees etc and we will cover them in future blogs. For the purpose of this article I am taking just the Employee Centric processes.

Employee Centric Processes – For those enterprises that care!

Today several enterprises understand how critical it is to let employees do their jobs! To do that we have to make every other part of their lives friction free as possible. So at Mahindra and Mahindra (winner of SAP Ace award for employee engagement) every employee has an AI assistant that helps them with almost every need of theirs from syncing calendars, to leave applications, from finance numbers to payslip downloads. The key is one AI assistant that can trigger or help humans execute over 76 enterprise processes! You can find the complete list of employee centric processes that have been automated on the NLPBots platform. The same assistant helps conduct dipstick surveys and instant notifications. It’s the holy grail of employee engagement!

Pre Covid 19 we thought that one of the most important reasons employees work in the office is that colleagues are just a few cubicles away. Being close to workers dramatically increases productivity. Working from home, employees cannot walk over to their colleagues’ desks. Hence working from anywhere but an office was considered impossible.

Today, however, employees working from home can communicate as seamlessly with each other as they could while in office thanks to collaborative platforms like Teams, Meet, Slack etc. The NLPBots AI assistant is available across all these channels from where employees can trigger these processes.

Broadly within employee centric processes we have the following types of processes

  1. Queries – these could be generic like “what’s the maternity policy in my company?” or specific to my band like “If I were to travel to Mumbai, what’s my daily travel allowance?”. These queries get resolved by the CPA doing a natural language search within the unstructured document repository (Corporate policy, etc). There can also be specific queries like “what’s the status of my expense claim that I raised for my mumbai trip in May?” or “how many vehicles did we sell in June?”. Here the CPA triggers a process that pulls data from a structured data source like Salesforce or Oracle.
  2. Transactions – These processes are either initiated via chat or voice in natural language or sometimes the human in the loop needs to react to push it down stream for further processing. Example “I want to apply for casual leave for the whole of next week” Here the system will have to a) understand that employee id XX is asking for casual leave from Date A to Date B and then b) confirm what it understood with the employee c) check it there is enough casual leave balance and ensure that its not violating any policy for XX c) inform XX if not or else inform approver if there is leave balance d) provide approver additional information along with the approval request e) understand and confirm the approvers action, etc. All the while it’s pushing and retrieving data from databases and various systems.
  3. Surveys and Competitions – Here the system triggers the process based on events or action. The responses are collected from the employees and pushed into a system or an excel and computed.
  4. Other processes – these include processes like authentication, notifications, 3rd party systems (employee mental health) etc.

Omnichannel and versatile

The versatility of the platform also extends to the channels of deployment and systems of unstructured and structured data sets. We have great IP when it comes to unstructured data, where you can upload and train your entire knowledge base as it is. No more manual intent creation and classification. This means that apart from creating better employee experiences at lower costs you can actually integrate data silos and start generating great insights. So far the NLPBots CPA platform has successfully integrated with various intranets, Whatsapp, HTML pages, Workday, Slack, Skype, Skype for business, Teams and various other chat and  collaborative platforms. The use cases and channels are not limited to the above.

When it comes to employees, they have a common pool of data (usually HR policies, Standard operating procedures, etc..) and some of the policies could be role specific. Based on the employees role in the organization they have various systems and data governed by different policies and privileges. With the NLPBots CPA platform you have a mechanism to add the data and workflows applicable to specific roles and create an AI assistant that is personal to each of the employees in the organization covering all the aspects of their work like. For example a IT security manager in a particular location will have access to data and work flows of general employees, data and works flows applicable to managers as well as his/her IT security role specific functions through a single window. You can adapt the workflows, permissions and data privileges in the organization directly on the CPA platform to create a enterprise AI assistant that helps every employee with Cognitive Automation.

All of these aspects and many more make NLPBots CPA Platform a really powerful tool when it comes to building cognitive abilities within employee centric systems. Of course this is just a tip of the iceberg. Imagine the possibilities when we open it up for integration with other systems and solutions out there. It means that any new vendor you onboard if they have a solution for your employees and a system with an API that service will also be available through the same window. And any other intent classification system out there will collapse as for those systems this will mean not only configuring the new use case but also going back and reworking the older utterances for previously classified intents as the confusion matrix will go through the roof.

Deep tech Meets powerful UI

This is only possible when you take superior NLP Algorithms, train with lots of data on Transformational AI neural networks and make it consumable on a great intuitive UI. So now OEMs, SI partners, Developers and even SIs can access the tech and create use cases that automate employee centric processes like KYC updation, Claims, Attendance regularisations, Approvals, Insights generation and many more.

It’s true, now more than ever before that the happier your employees are the better the yield. So as organisations are battling productivity issues while getting their employees back to work to produce and sell their products or services they’ve got to stay motivated and productive. For employees it means removing all hurdles to their performance. When the systems are able to take care of the mundane, repetitive tasks they do that require them to do it just because the enterprise systems couldn’t, it frees them up to do the real thinking. It helps sales teams to sell better, HR teams to empathies with employee problems better, Operations to optimize resources better and the entire enterprise to deliver more. Cognitive process automation platforms help the employees “see” better, “read” better and act with decisiveness on their core tasks. The classic case of employee – AI collaborating to achieve much more.

So if you are looking to deploy an AI workforce that caters to your employees do get in touch and we’ll show you how you can create impact within the week.