Medium and enterprise organisations are currently looking inwards to solve internal business problems and streamline internal processes.
We have built solutions that automate HR responses, centralise communication and documentation and are a 'central point' knowledge system (a chatbot that knows all company information).
Chatbot technology is evolving rapidly, and they are a viable front-of-house communication tool, so, why are enterprise chatbots mostly internal?
Well, there are three main reasons.
Risk, testing and systems integration.
Despite being decades old, natural language processing and machine learning are relatively commercially-untested technologies. Up until recently, they were more commonly discussed in academic research papers rather than on Medium.
Lack of best practices, case studies, past experiences and evidence leads to a higher risk, and if there is one thing most big companies do not like, it is a risk without known reward.
For the innovative and forward-looking staff at larger companies, it is tantalisingly frustrating. Those that have taken the plunge have started to show AI-driven conversations can solve real business problems, increase revenues and compliment customer service.
They want to get in on the action; they want to implement the tech. But how do they do it without upsetting customers, risking sales and service revenues? Well, they find a business case and implement internally first.
As an example, our ubiHR chatbot is for those companies. It is a conversational solution that lowers the burden on an HR department by instantly answering some of their inbound questions and information requests.
One of the companies we are working with has an HR team of four people dealing with 1,500 enquiries per month. How much annual leave do I have? How do I claim for XYZ on Bupa? What is the company paternity leave policy? What does this mean in my payslip? We reduce the number of these types of inbound communications by 30%-50%. That HR team just got back an extra two weeks for every month.
By implementing internally first, organisations can understand how to apply AI technology, its impact on business processes and, perhaps most importantly, how to build a business case around it. All done without anything consumer-facing.
Language, persona, brand, features, functionality, channels, infrastructure, support staff, internal training, vendors and APIs. Just 11 top-of-head things that we need to test when developing and integrating chatbot technology.
Broadly, there are two options to deal with testing.
Internal and external
We do lots of internal testing of a chatbot before handing it over to the customer so that they can do a whole lot more. We think of everything that can go wrong, we think of all the questions it might be asked and all the different answers it might give. We test message receipt, delivery, API endpoints, infrastructure, software, redundancies and everything else. Between the customer and us, a whole boat-load of testing gets done.
External testing is pretty much the same thing, only this time we let strangers loose. We do structured testing like "please test resetting your password through the bot", or, "imagine you have just found out you are pregnant, please speak to the bot about it". Whereas unstructured testing means we let the audience loose without any instructions or guidance.
There's no doubt about it. A chatbot deployment needs to be tested thoroughly (and we are not even going to touch on machine learning here).
Thinking about the company again, with this new technology and amount of testing required. What do you think they would prefer? Testing unknown technology on their customers? Or, with their staff whom they have better communication channels with, more understanding from and, arguably, a lesser impact on revenue if it does not work as planned?
Boy, do big companies have many systems. CRMs, CMSs, supply, sales, and marketing software, HR, benefits and payroll systems, analytics, reporting and monitoring suites. And they are only a few I could think of as I was writing.
Do you think they want to take a new, risky and untested technology and start trying to integrate into all of this? Heck no. Not on your nelly.
Even if you have the most tested, time and money saving/making software on the planet, in the enterprise space, more often than not it is integration time and cost that will ruin the deal.
Consider our ubiHR example from before. To get that product up and running, a company has to do two things. The first, send us the information they want to be turned into a conversation. Employee handbook, benefits package information, relevant policy documentation, etc. (it is all secured and encrypted btw). Second, they put a line of code in the header of their HR platform, CRM or any other place they want the chatbot to appear.
That is it. This type of low risk, testable and integration-free solution is the exact reason why enterprise is looking inwards first.
The types of inward solutions they are implementing? Well, here are a few examples:
24/7 online and on-demand HR service to answer employee questions on leave, payslips, salary, benefits, maternity and paternity and more.
Real-time accounting reporting and data gathering to speed up expenses, payroll processing, invoices and settlements.
Centralising of documentation and information to ensure everyone has a single point of up-to-date company knowledge.
Internal sales tool for product information, supply chain mapping and sales advice.
Internal customer service tool to find product specifications, sustainability and sourcing information.
Central analytics, reporting and governance with user role restrictions for company-wide information gathering and display.
Meeting and appointment scheduling for displaced teams to arrange times to chat, work and manage projects.
Social media reporting, analytics and management.
So there we have it, the reason why enterprise is currently mainly looking at chatbots as an internal tool. However, trust me, they will be using them more frequently in externally-facing sales and customer service very soon.