There is a lot of interest at present in bots to provide a conversational interface that will integrate data from multiple systems. Some of my search projects have in effect created bots in the development of search cards. A common implementation of a search card is in a professional services firm (PwC comes to mind) where a query for a specific customer brings up the current status of engagements, client representatives, meetings that have taken place or are planning with the customer and often some basic customer information. Sam Marshall has written a good post on search cards and on the benefits of presenting information rather than results. If all the information required is on a single application such as SharePoint or can be accessed through a web call then the work involved in creating these cards is not too difficult.

Life is much more difficult when there is a need to integrate one or more enterprise applications. Enterprise integration planning has been on the CIO’s to-do list since the mid-1990s and it is not getting any easier. The range of enterprise applications only becomes obvious to an intranet manager when they look at the search logs and find that a significant number of the queries are generated by employees trying to locate an application either by name or purpose. IT Departments spend much of their time maintaining a very significant number of applications which are only used by a comparatively small number of employees. Several years ago one of my clients, a mid-sized global engineering company, found that over 700 different applications had been searched for in the course of a six-month period.

The concept of a bot that works across a set of applications to deliver information is very attractive but in the articles and columns I have seen no account is taken of the challenges IT will face in responding to what seems a simple request. This is especially if the company has grown by acquisition and it is easier not to change out some the applications for what may be notionally the master corporate application. That brings me to master data schema. If you have problems defining metadata for an intranet remember it is usually a couple of orders of magnitude easier than agreeing the elements of a cross-enterprise master data schema. If the metadata is missing or incorrect in an intranet context the intranet will still work. Missing data elements in an enterprise database can bring it to a halt. To make a bot work you may need to align the master data schema to the intranet metadata schema and that is a decidedly non-trivial problem. It is also important to remember that what looks like a web application might actually be a complex enterprise database with a web front end.

My advice would be that if you have some good business cases for bots you start talking to the IT Department at the earliest opportunity. It could be that a bot (or search card) has such potential value that it changes the priorities within IT as far as application integration is concerned. However, I suspect that will be the case only in a very small number of instances. The rate of development of bots does not depend just on the vision of intranet and digital workplace managers. Delivering the bot experience could require a significant effort by IT and someone is going to have to pay the bill for it.

Martin White