Are you seeing multiple potential chatbot use cases within your enterprise? Here’s how to take a consolidated approach to enterprise.
Let’s look at a common scenario that an enterprise faces with chatbots.
They start off deploying a pilot chatbot as a trial for a single use case. Then, on seeing the benefits, other departments and business owners within the same organisation explore their own chatbot projects. They often partner with different vendors they select on their own.
Soon, there is a proliferation of chatbot deployments by different business units in different regions. They have around 15 initiatives underway, 5 separate independent existing chatbot projects, and at least 10 different chatbot vendors deployed. After a while, this leads to redundancies, increase in complexity and inefficiencies in governance.
Does that sound familiar to you?
Then it's helpful to start thinking about a consolidated enterprise chatbot strategy. Here's a simple roadmap on how you can go about it.
Chatbots deployed for one specific use case can be considered for lateral growth, across other departments within the organisation. This will require extra attention on governance, content control and costs to match the increase in complexity as the use cases grow.
To achieve this expansion, you will have to think about features like access rights – who can see what aspects of the chatbot – and maker-checker which allows for review and approval by different stakeholders before bot tasks get published.
If you have multiple stakeholders in your enterprise, think about a horizontal scaling of the chatbot to serve their needs. These can be for internal or external stakeholders. At this stage you will need to consider features such as authentication and personalisation.
Think of a chatbot deployed for an insurance firm. It can serve various stakeholders ranging from external prospects and policyholders to internal employees, agents, distribution partners and contact centre representatives. Being able to identify users, and rules for responding conditionally come into play here. For example, if a query is posed by an existing customer, the chatbot should provide a personalised answer and if not, the response should be given with that knowledge.
Chatbots often start out as a way to automate routine responses to queries. In fact they may just be an interface to collect customer queries to be directed to a live agent to take on. But if you don’t integrate this bot with internal systems, you are missing out on some significant benefits.
When expanding your single use case chatbot across the enterprise, you will need to plan for integrations with existing systems. This is especially important for the chatbot to serve different stakeholders with the knowledge of their history and type of usage.
Take the case of a healthcare chatbot. It will need to work with the hospital’s electronic medical records, medical and drug databases to serve patients and clinicians respectively. Without these integrations, the bot may still be able to respond to general queries. But it will not be able to help provide personalised responses to patients, automate doctor appointment booking and other value-adding tasks. The hospital’s sales team could also be using a CRM tool to manage leads and customers. Integrating with the CRM tool can help the bot personalise responses to convert leads into customers more effectively.
4. Add more channels, platforms and plan for multimodal chatbots
The mode of communication in the chatbots doesn't necessarily have to remain in the same web app interface. It doesn't even have to text-based. It is helpful to have chatbots expand across channels and platforms to cover more users and use cases.
A common scenario for example, is a chatbot initially deployed on a standalone portal expanding to WhatsApp, Facebook Messenger, Telegram and other instant messaging channels. Adding other forms of input, such as voice, can also increase the chatbot's usage.
If your organization has a presence in different regions, there's no reason why you can't expand your chatbot's usage across these geographies. This will of course mean accounting for differences in language and dealing with additional complexity in managing different knowledge bases.
Multilingual chatbots are those which can take input from users in different languages and produce the appropriate responses. These often involve NLP models that can employ cross-lingual transfer, exploiting the fundamental similarities in the structure of languages.
You will also need to plan for compiling the necessary language datasets and adding more examples of real-world queries. Look for vendors who have prior experience in deploying multilingual chatbots and who can help you centralise the knowledge bases and management of the different languages.
This step can sometimes be the first in the expansion. The goal here is to ensure a seamless handover from the bot to a live human agent. Essential features to get right in this stage include:
The last part is crucial as there needs to be a mechanism for the bot to learn from the live agent what can be done in the future. For example, the agent can tag each of the chat texts from the user to a corresponding intent, so that it appears as a new variation of the selected intent.
As you expand your chatbot use cases across the enterprise, how will you know which ones are worth continuing and which ones are not working? This is where analytics is important. Make sure you have the analytics tools and metrics in place to measure the performance of your chatbots, especially across channels, personas, departments and countries.
KeyReply can help you take your single use case chatbot and scale it across your organisation to be an enterprise-wide chatbot. To find out how we do it, ask us for a free demo today.