Chatbots may seem like just another tech buzzword, but they actually have the power to transform how businesses work with data in a big way. Instead of companies investing months of effort into building complex dashboards and BI tools, chatbots let users have conversations directly with their data from day one.
Chat interfaces can rapidly empower employees to slice and dice enterprise data on their own terms by focusing the conversation on the insights that matter most, not the limitations of a rigid, predefined interface. You can think of chatbots as your personal analytics assistant, available anytime to fetch relevant data and answer questions as they come up.
Conversational UI is a great way to quickly access a company’s data. Take numerical data, for example. Rather than forcing users to rely on a suite of BI tools like Tableau or Google’s Looker or building custom dashboards, chatbots offer a single, flexible interface to access data. With rigid tools, data analysts first need to figure out the most critical metrics and signals, and then a team of designers, engineers, and analysts build features to present that data.
Chatbots shift the focus from building a standardized UI to making the data itself more accessible. With a chatbot, users can instantly decide which data points are most important and get that information. The chatbot is a conversational assistant to slice and dice data on demand.
Chatbot for a Marketing Agency, An Example
Let’s look at an example. If you’re building a CRM system for a marketing agency, it would typically have a database of agency clients. Each client would have a set of meeting notes created by account managers along with other client-related data. When you meet with a client, you look at the previous meeting note, review the performance of the client’s campaigns in a standalone tool like Google Analytics or Tableau, and prepare a report for the client.
At some point, this work becomes repetitive, and an agency might want to create a dashboard in the CRM, if it’s custom-built, showing the performance of the client’s campaigns so that the managers or other team members don’t have to collect the data in several different systems manually.
With a chatbot, you can skip the dashboard step, at least initially. Instead, you can sync the data from analytics and BI systems with the database a chatbot has access to. Users can then talk to the data directly and ask questions like “How are the campaigns for client X performing?” and get a list of campaigns, number of leads, and conversion rate per campaign. Users can even ask more detailed questions — for example, “I have a meeting with client X today. What should I talk about with the client?”.
If built on a framework like LangChain, the chatbot can have a tool that supports this type of prompt. It would fetch the last few meeting notes, campaign performance data for this client, maybe some additional website analytics from landing pages and overall sales numbers and give you an extensive list of things to report to the client, questions to ask, and suggestions of what the agency can work on in the next month.
In the example above, the chatbot is the interface. You can skip building all that functionality in the main CRM application and instead return the answers in a textual form, which is way easier and faster to implement. It saves you a ton of effort on the development team — no design, backend or frontend engineering required, low or no code. The chatbot provides a natural language interface to the data.
Chatbots Allow Rapid Iteration
The process of building an AI chatbot is much faster and more efficient than traditional development. Plus, you get to collect the requirements much faster. Instead of spending hours meeting with stakeholders, iterating on mockups, and gathering feedback, you focus only on the data they need and how to present it meaningfully. Changing text is easier than rebuilding a UI.
Compare that to traditional dashboard development.
First, you’d collect data requirements like with a chatbot. Then, business analysts would work with stakeholders to determine interaction and presentation. Should it be tables or charts? Do they want campaign comparisons? Historical data? What existing systems are they used to?
Next is design and development. The design team prepares mockups for stakeholder review. Then high-fidelity mockups and prototypes. The engineering team collects data and builds backend APIs. Finally, the frontend team builds the UI for desktop and mobile.
That’s far more work than a chatbot. With a chatbot, you collect the data, build the tool, test it with example prompts, and deploy it. It won’t be perfect at first, but you need to collect the data anyway, regardless of the UI. Building a conversational UI is faster than full-featured dashboards. The best part is you can rapidly improve the tool based on user prompts. It’s much easier than traditional analytics and feedback.
And that’s just one use case. Integrating a chatbot into directly into a custom-built CRM system makes it even more useful because it can have the context and access to company data. Ask anything about a client, contact, or meeting you’re looking at. It’s like having a capable assistant ready with data and answers.
Then, you can make it 10x more useful by embedding widgets in chatbot responses. Ask about a campaign and get a response with a “Learn more” button. Click for a generated report right in the chat. The possibilities are endless.
Combining Chatbots and Traditional UIs
Startups promising to let companies talk to their data are popping up everywhere. But like company-wide search, these chatbot features need tailoring to each business and integration into internal systems to fully realize their potential. They can help get you started, but general-purpose products inevitably hit limitations.
Chatbots are a great way to quickly empower users and enable them to query data directly. They allow engineers to focus on collecting and processing the data, not building UIs. But as you add more data sources and tools, the system grows more complex and harder to maintain and improve.
Chatbots also have a caveat — users tire of repeating questions and forget what to even ask. To be truly useful, chatbots must go beyond responsive; they must proactively flag potentially useful info and suggest prompts to guide users.
The best solution combines chatbots and traditional apps. With chatbots, you can rapidly prototype and iterate, delivering value quickly. But once a use case proves itself, confidently integrate it into the traditional UI.
Originally published on Medium.com
