10 Tips how to get started with Chatbots in an Enterprise

Anytime I talk to a larger organization I get the question: “ok, we want to start with chatbots, but how would that work for us?”, “What does it take to build an enterprise-grade chatbot?”.

For any smaller business or startup, it’s easy to get started, but enterprises have a lot of politics and legacy to deal with.

Here are 10 useful tips—based on real-life experience—to help you get started with bots in an enterprise environment.

1. Define a business case

Start with a business case. Write down what the bot should achieve, what it’s KPIs are and what the ROI is.

I’ve written another post explaining this in detail, make sure you have it at the start of your project.

2. Find a business owner

This is a simple, but vital step in any chatbot project that wants to run in production. You will need a business case and a business owner, a champion, that will support your project.

This is not you, but someone sharing the same interest and has a decision making role. Your business case should point out who this could be—based on shared KPIs — or at least help to find and convince the right person.

Some examples of business owners are:

• Manager customer care
• Manager customer experience
• Manager of online marketing
• Product manager
• C-level executive

Even for innovation projects, you will have a difficult time—even after successfully running a pilot—to deploy it for everyday business if you do not find a person moving it through the organization’s governance.

3. Create an assumption

It’s part of the business case, but if you choose not to create one, write down at least an assumption with an estimate.

Some simple examples:

• Using a chatbot we can resolve 15% of customer service inquiries
• A marketing campaign on Facebook will achieve a 300% higher engagement rate than our existing email campaign
• Using a proactive chatbot we can increase the lead generation on our website by 15%
• We can provide an additional service for our customers using a chatbot on WhatsApp and our customers give it an average rating of 8 out of 10

Create an assumption you can easily test or measure during or after deploying a chatbot.

After going live with your bot, you will want to see how successful it is.

4. Start with an MVP

Chatbots are proven technology but they might be new to your organization. Start with a pilot to validate your assumption. For that, you do not need a complete end-product but a minimum viable product (MVP).

…you do not need a complete end-product

Avoid making it a huge in-house IT project in the early stages.

Trust me, you’ll waste months of time and money before validating any of your assumptions. Even if your IT department is very eager to do it themselves.

Instead, try to keep everyone involved but push an MVP live within a few weeks.

5. Test your bot internally

With usability testing, just a couple of persons are enough to gain insights. One of the benefits of working for larger organizations is that you have a lot of colleagues.

Bonus tip: I often experience internal stakeholders do not test until the bot is actually live and then provide feedback. You can fix — or cheat on that — by telling everyone your bot is live on a certain date, but you will actually do that a week later.

Testing a bot internally will not only give you insights into how end-users interact, but it will also grow internal support.

6. Test it with customers

After testing your bot with co-workers the best way to validate and testyour bot is putting it live-in front of customers. Bot interactions give fine-grained insights on how customers interact.

It’s not bad to see your MVP is a total failure

Can’t do it with real customers? Find or create a test group. The most important thing to do is validating your assumptions.

An internal test with colleagues will help to find any bugs or quirks and validate your product.

It’s not bad to see your MVP is a total failure. Better now than later. Remember that it’s why you work with an MVP in the first place!

7. Validate your results

After testing the MVP, preferably with real customers, you’ll gain insights that you can validate:

• Is your assumption correct
• How do customers interact with the bot
• What features do they reed
• What must be done to deploy and run the bot in production

It will also give you a clear picture if you are ready for prime time, or what is needed to get there. All this information can be used for reporting.

8. Test it again

Because you have an MVP it is easy to make some simple adjustments. It isn’t done often—I confess—but if you have the time and some budget, it would move your MVP a bit further while re-validating the last pitfalls.

See it as a revision you could also plan in between your project. Planning could look like:

• Test the bot internally
• Revision 1 based on feedback
• Test the bot with customers part 1
• Revision 2 based on customer feedback
• Test the bot with customers part 2
• …

9. Create a stakeholder presentation

If you continue with your bot, create a keynote with a couple of slides to present internally.

In larger enterprises, you will continuously need to convince stakeholders, even after deploying the bot to production.

…you will continuously need to convince stakeholders, even after deploying the bot to production

Such a presentation can contain the following topics:

• Initial business case and (or) assumption
• Business owner and stakeholders
• Summary of creating the MVP
• MoSCoW or other some other method describing the choice for MVP functionality
• Measurements and test results
• Security, data privacy and choice of product/ technology explained
• Next steps needed to deploy the bot, think of the budget, agreements, etc.

10. Move it to production

So you successfully proved your bot adds value. How do you move the bot into production and everyday business?

Unfortunately, a lot of bot projects end here.

No business owner

Without a business owner, I’ve never seen a bot ending up in production. You need to find someone to be your internal champion.

The bot should help in achieving their KPIs and they will help set up a governance model and convince stakeholders inside the organization.

Department X disapproves

It’s often hard for business units, like marketing and customer service, to explain how their bot helps in achieving KPIs. They have to deal with IT and procurement departments and those usually have their preferred suppliers.

For example, IT departments traditionally like to consolidate tech into single-vendor solutions. This is understandable, but it can stand in the way of successfully deploying bots to production.

A solid presentation, business case, and KPI driven results will help a lot in convincing stakeholders.

Involving IT, procurement and other departments from the start will NOT fix this problem. In my experience, it has never affected the outcome, having a stakeholder presentation, valid KPI based results and business owner does.

Unless you have enough leverage higher up the chain, you’ll run the risk losing all momentum, either turning it into a big IT project that — trust me on this one — will not deliver on any of your business KPIs, or will silently fade away.

About The Author

Gijs van de Nieuwegiessen is CEO and Co-founder of Flow.ai, a Chatbot design, and creation tool. He has a background in UX design, a creative technologist, worked for large enterprises and started his first company in 2005. You can connect with him on Twitter or here on LinkedIn.