[Skip to content]

FM World logo
Text Size: A A A
22 March 2019
View the latest issue of FM
Sign up to Facilitate Daily >
FM World daily e-newsletter logo



Chatbots for FM-specific functions are still in the pilot phase but they could provide value as their capabilities mature, explains Sam Marshall.


06 June 2018 | Sam Marshall

A lot of what chatbots do is add a natural language layer on top of a search engine. So if your search enterprise struggles to bring back results accurately owing to badly indexed content, a chatbot won’t fix it. 

It’s going to distract you from getting the fundamentals right.

The most basic chatbot uses a table of questions and answers (like FAQs) to provide a piece of text to answer a query, but there are smarter bots. 

AI and true conversation

A typical conversation with an FM helpdesk might involve a user requesting a meeting room for an hour. The helpdesk person says: “Yes, room two is available.” The user asks: “How many people can that hold?”

It’s a simple question for a human to answer, but a chatbot that treats each input as unique would say: “Sorry, I don’t know what you mean by that.”

Even consumer-grade chatbots with millions of investment in them are not that conversational. 

AI-capable chatbots are an improvement on these basic bots because they are able to maintain that thread and all these kind of indirect references for a smooth transaction.

Machine learning

The third level of intelligence is machine learning so if, for example, 20 people on the same day ask the chatbot where they can get a temporary pass, the bot will detect the pattern and prioritise that answer in the same way that Google does with frequent questions. 

The bot can also learn new terminology, so if people had been asking how to set up a WebEX and then wanted to set up a Skype meeting, the chatbot will understand that the request is for Skype with web-conferencing and not a consumer Skype chat. But few bots are ready to run for an FM to pick up and do that.

Six prerequisites for successful bot integration

 Bots should be available in the interface people already use to avoid having to access the web or open another application.

 Bots should be maintained by the content owners to remove delays with third parties having to make changes to content. 

 Bots should be transactional, so as well as being able to tell you how many projectors are available, it must also be able to book them. Think of bots as replacements rather than searches.

 Bots should operate across company departments to remove multiple steps to determine where the request should be logged. 

• Bots must provide answers, not links. 

 Bots must be capable of true conversations. Chatbots should be able to cope with typos, as well as slang, and abbreviations that people use in normal speech.

What’s a chatbot good for? 

Anything that uses a simple form of two or three questions to complete is easy for a chatbot to handle, such as a lunch booking, with the date, time, size and dietary requirements of the party. 

Bots could add value by negotiating around resources; they would be able to swap Sally’s booked meeting room with John’s because Sally has more attendees. That is a lot of time for a PA, but there is no issue for a bot to sort it out. 

Hot-desking presents a challenge when trying to locate people in the workplace. A chatbot could tell you where a person is sat. Bots could also handle maintenance requests and data about car parking permits.

The appeal of bots for FM

Bots could, for example:

 Cut down the cost of service provision as it takes on calls to the helpdesk;

 Guide users through all of the information, eliminating a back and forth of emails; and

 Allow staff to request FM services without having to find a specific application form, as the bot does it for them; 

There is scope for bots to become more like Google – to advise, for example, taking a train in two minutes to get to a meeting in an hour. Why don’t we have bots that can tell you on your way to work that car park A is full so you’d better head to car park B? Those would be most useful. 

Sam Marshall is director of ClearBox Consulting