Work Simple
The diagnose first approach
Many field service organisations today still use a “diagnose first” approach. Typically, upon receiving a call for an incident, an IoT failure message or a service sales order, the organisation prefers to first send a general assessor out in field. This initial agent’s responsibility is to understand the nature of the failure or service to provide, generate an estimate of the work to do and return to base. The estimate is then reviewed, approved and an agent with appropriate skills and parts is dispatched to provide the service or resolve the incident.
There are a few reasons why organisations proceed in this manner. The most obvious one is to ensure the work plan is appropriate before carrying out any work that might otherwise incur undue costs or further failures. This is often due to a lack of information to ascertain the nature of the problem.
There’s also a cost saving rationale behind this. It’s more economical to have a small number of highly specialized agents (as opposed to all agents having all specialities) and to allocate the right agent with the required skill set to the job.
That argument is especially true for rare or highly sought-after skills. If only 2% of your total monthly expenditure is allocated to fixing a specific failure that requires an expensive skill, you want to avoid paying for 100% of your workforce to have this skill. You’ll most likely want to use a blended model workforce, leveraging specialised contractors for this specific failure and send the latter only when truly necessary.
So how do you solve these 2 problems without having to carry out multiple visits? With the information at hand at the time of the request, how do you ensure the work plan is appropriate and you are sending out the right resource with the right parts that will resolve the issue the first time?
Phase 1: Disconnected systems
When systems are not fully connected and sending out an agent to site means losing contact for a few hours or even days, the solution is to rely on statistics and provide agents with a level of decision authority.
Statistically, most failures fall within a range of possibilities. Workforce is recruited by roughly matching skill set and expertise to resolve these failures. At the time of dispatch, a bet is taken as to where the failure occurs within this realm of possibilities.
In field, the agent diagnoses and then assesses whether he has the skills and parts to resolve the issue. Moreover, he is given a tolerance envelope: Can the work be done within this budget? If the answer is “yes” he proceeds. If statisticians are correct, the cost of an assessor is saved for approximately 60% of incidents or events. The remaining 40% requires a second visit.
Phase 2: The connected solution
As service providers extend their offering, asset ranges expand and skillsets become fragmented, no field technician can know everything. This reduces the possibility of finding this “60% baseline”.
To some extent, this problem is solved with the use of real-time connected field service management systems. Documenting the proposed work plan, reviewing it, correcting it and approving it can be accomplished in real-time, while the agent is still on site. Based on rules and decision trees, the system can approve a work plan and immediately generate a revision of the current work order which is then relayed to field personnel in real-time.
This in turns expands the agent’s diagnostic skills and gives office-based managers the confidence that work is being supervised by a central and predictable set of rules.
Skillsets can also be enhanced during execution of the fix per se. Having had the chance to review and approve work plans, the system can attach specific documentation, videos, knowledge articles to work orders, and automatically send these to mobile devices. The field agent can refer to the latter to fill knowledge gaps or ensure he is following recommended protocol. This is also where real-time chatbots or “enhanced reality” would come into play.
This real-time feedback loop between the field and the connected system means that you can make better use of the “60% baseline”. Agents can now get immediate approvals and be up-skilled in real-time during their initial (and only) visit. Previously, they may have had to pack-up and be replaced by someone else.
Phase 3: The connected AI solution
Despite this, we are left with hard physical constraints like missing spare parts or required regulated skills. Even with the best system guidance and approvals, an agent missing parts will need to request for a second visit.
The best solution is always to have all requirements and information beforehand. However, the way in which failures are reported (from the public, from external observations, etc.) often means very little information is known at the time of dispatch.
This is where AI comes in. Machines are much better than people at detecting patterns, especially when using large numbers of inputs that we humans might not always consider.
As an example, a request comes in for a temperature control problem in a commercial building. A traditional system may classify this as an HVAC work type and dispatch an HVAC specialist to first do an investigation. In the last few hours, following a very hot day, 2 similar requests have been logged for buildings belonging to the same owner, but having different HVAC manufacturers. In both cases, the root cause was an electrical distribution problem due to an original faulty installation. The AI powered system extrapolates this information and proposes a work plan for an electrical fitter instead of an HVAC specialist. This provides the highest probability of “first time resolution” and avoids the cost of a useless truck-roll.
In the above example, the service provider doesn’t need to leverage the connected review and approval or real-time support capabilities. By letting the AI consider all possible inputs and extrapolate patterns, it sends the right technician with the right skill the first and only time. Moreover, it optimises time in field by pre-emptively providing full contextual knowledge of the issue prior to dispatch. Impressively, it’s able to do all this with only an initial basic description of the problem.