We’ve helped a lot of businesses deploy chatbots. We’ve seen it all.
When done right, business chatbots are one of the best ROI initiatives your company can invest in. But too many inexperienced companies run into the same pitfalls when it comes to deploying their chatbot.
After several years and thousands of chatbot deployments, our Customer Success team compiled a few of the most common mistakes they see organizations make with chatbots.
Here are a few of the most common – and easily avoidable – mistakes that companies make when deploying chatbots:
1. Tunnel vision for a single-use case
HR chatbot. Customer service chatbot. IT chatbot. Hotel booking chatbot.
Most chatbot initiatives will start with a single use case. Your team figures out what works and what doesn’t, and then they deploy. That’s perfect – but once you find a recipe for success, don’t forget to keep repeating it.
As long as your chatbot initiative isn’t limited to a single use case, you can use the same software to automate processes across company departments.
If you want to get the best ROI out of your chatbot investment, keep repeating the process.
Start with an FAQ chatbot for real estate and expand it into a lead generation chatbot. Or start with an e-commerce chatbot and then build an HR training bot.
Their endless use cases are why chatbots are a cost-effective investment – don’t neglect your chatbot’s potential.
2. Forgetting KPIs
How can you tell if your chatbot is successful if you don’t pick the metrics to measure its success?
When scoping your project, you need a sense of what you want your bot to accomplish. Is it going to divert calls away from your overburdened customer service reps? Is it going to capture leads by dispensing industry information?
Your KPIs must be:
- Concrete numbers
- Backed by existing data
- Possible to measure (with a plan for who will measure them, how they’ll be measured, and when they’ll be measured)
Before you embark on a chatbot project, you should have your KPIs in place. They may evolve, and that’s fine. But a company needs KPIs in order to properly build and measure a successful chatbot.
3. Tasking an intern with a full chatbot deployment project
While your latest batch of summer interns will surely feature a few raised hands to take on a solo chatbot project, we’ve got news: it’s a demanding project.
A chatbot is a development project, just like any other software. There’s no quick copy-and-paste solution that’s suitable for a real enterprise to put out to their users. A company needs long-term employees ready to see it through.
Our Customer Success team recommends that a chatbot project involve 1-2 developers and 1-2 business-side employees. That one coding class you took in college probably isn’t going to cut it.
4. Not balancing business and developer needs
Solutions like Langchain are perfect for developers. But it means team members on the business side are usually unable to collaborate on deployment.
Some of our competitors – we won’t name names – are perfect for business decision-makers. But once they hand it off to the rest of the team, their developers’ hands are tied by limited platforms.
A chatbot is a collaboration between a developer team and go-to-market team. A successful deployment is a marriage of both. Make sure your roadmap and tools are well-suited for both sides of the equation.
5. Underfunding
“I’m looking to build a chatbot for $200.”
If you want a chatbot that’s going to add tangible value, keep looking. We’ll say it again: a chatbot is a real software development project.
The price of your project will depend on your scope, needs, and dev team. But be prepared that an enterprise-grade chatbot is going to cost you a lot more than $200.
The good news is that if you deploy it properly, you’ll easily see the upfront price of your chatbot return in ROI.
6. Ending the project after deployment
So your chatbot is out in the world – project done, right? Absolutely not.
Plenty of companies see a chatbot as a one-and-done process. They pay, they build, they deploy, and they’re done. But if you’re trying to get real value from a chatbot, deployment is just the beginning.
A chatbot needs to be monitored. Reviewing, feedback, adjustments, and iterations should all be planned for in advance.
The software lifecycle includes iterations of testing and maintenance – and that includes a chatbot.
7. Not building new processes
Chatbots should be built for a reason. They should be inextricably intertwined with your business processes – not a give-or-take side project.
If you don’t build new processes around your chatbot, what’s the point? You won’t see anything close to the levels of ROI that chatbots are capable of.
At Botpress, our chatbot is the first level of customer support. No matter the question, everyone begins with the bot. If needed, they’re transferred to a human. But remove our chatbot, and our processes need to change – it’s an integral part of our pay-as-you-go customer support process.
While scoping your project, make it clear what your chatbot is going to automate and what new processes it will bring with it.
8. Overselling to upper management
It’s easy to get carried away when pitching to upper management. But overselling upfront is a recipe for disaster later on.
For example, industry standards for NLU understanding rates used to hover around 75-78%. But if an enthusiastic associate pitched a 95% understanding rate to their boss in order to get project approval, then even a wildly successful 85% will look like a failure.
Give reasonable expectations. Half the clients we see oversell upfront, which only leads to disgruntled management down the line.
And if you’re not sure what’s reasonable, just ask. Our Customer Success team is always happy to sit down with clients to go over reasonable expectations to present to decision-makers.
9. Leaving 100% of the conversation design to LLMs
“I’ll just leave the conversation part to ChatGPT.”
Just because your company’s project will be powered by an LLM engine doesn’t mean you should neglect conversation design.
While a good chatbot platform will simplify best conversation practices, companies need to be aware of their unique user experience.
LLM agents can provide a natural conversation flow, but bespoke conversation design is necessary to perfect your user experience. Just because you're using agentic AI doesn't mean your bot will fulfill your bespoke company communication parameters.
No conversation design experience on your dev team? No problem. We’ve got a Botpress Academy class on Conversation Design for new bot builders.
10. Short timelines
An optimistic client of ours recently told his boss that he would have their chatbot up and running in 2 weeks time. And you know what? He almost made it.
A standard timeline for a chatbot – if you’re trying to expedite the process – is 1 month. Any shorter than that, and you’ll run the risk of a) missing your deadline, or b) deploying a bad bot.
While the building process won’t last a month, you should accommodate the extra time to account for testing and reiterating your bot’s structure. It’s always better to have extra time than not enough.
11. Trying to get value from a no-code solution
There are plenty of chatbot platforms on the market. And a lot of them advertise themselves as ‘no-code’ solutions.
But to deploy a useful chatbot, you need to get technical. Anyone who tells you otherwise is trying to sell you on a limited solution.
Sure, you can use a no-code solution to build an FAQ bot. But what value is it really adding to your business?
High-ROI bots are integrated with your company’s systems and day-to-day workflows. Building a useful, seamless integration to your company’s bespoke business processes will – don’t be scared – involve some coding.
No-code solutions are fine for basic bots. If you’re building an enterprise-grade bot that you want to see returns on, you’ll need to enlist the services of your dev team.
No developer on your team? Ask to connect with one of our partner organizations – we’ll match you with a bot-building expert.
Building a bot? Do it right
If you’re going to build and deploy a chatbot, do it right.
When properly built, a chatbot can be the highest ROI project on the table. But if your company runs into a few of these common mistakes, it can be a frustrating project with murky value.
We’ve spent the past 7 years deploying chatbots to enterprises. We know what works and what doesn’t.
If you’re looking to deploy a high-value chatbot, we have the experience to set you up for success.
Start building today. It's free.
Table of Contents
Share this on:
.svg)
.webp)
