How AI conquers your planning challenges in operations

AI tackles planning challenges in operations with an optimisation engine that weighs every job, technician, travel time and SLA deadline at once. It updates the plan as soon as the day changes. The planner sets the rules and values the engine works with, and decides on the exceptions.

In a previous blog, Erry de Boer, CEO of Eqeep, set out his view of the impact of AI in finance. Here, he shares his practical insights on operational planning, and we show how the IFS scheduling engine works.

Why is your schedule out of date by nine o’clock?

Because the day moves and the schedule doesn’t. Planning your people, machines and vehicles requires continuous attention. Even rescheduling around SLAs is still largely manual work at most companies. When an urgent call comes in, your team must work out who is free, who goes where and which deadline applies.

In Erry’s view, manual planning is time-consuming and increases the chance of errors. Heads of planning tend to recognise these problems:

• This morning’s schedule is out of date by nine, because a job overran or a technician called in sick.
• Only one planner knows what takes priority, and when she is away, things stall.
• The schedule lives in a spreadsheet or on a board, and an emergency means phoning round.
• Nobody can work out what a large new contract or two departing technicians would mean.

How does AI take away your planning headaches?

By setting every option side by side within seconds. The system combines all the important data: the content of your SLAs, the availability of your people, their skills and customer knowledge, and stock. If you have recorded all of this properly, AI suggests what you should do.

Take an emergency repair for an important customer in The Hague. Technician Henk is on his way to Amsterdam, but he has exactly the right expertise and the required parts in his van. The AI system analyses all the options at once and sees this:

• Technician Henk can be in The Hague within an hour, because traffic is light.
• Technician Klaas, who also has the right expertise and stock, will finish his current job in thirty minutes.

The system adjusts the schedule. Klaas takes over Henk’s appointment in Amsterdam and arrives exactly on time. Henk drives to the emergency job in The Hague.

In IFS PSO, this swap goes through the planner or through commit rules you define. Henk’s job was already assigned, so the engine would not move it alone. The planner sees the suggestion in the Dispatch Console and decides.

How does the engine know which job comes first?

From the values you set yourself. An optimisation engine has no idea what ‘important’ means. It weighs the value of each job against its cost, and looks for the schedule with the best total.

In IFS Cloud with IFS Planning & Scheduling Optimization (PSO), the IFS documentation describes it like this. Each job gets a base value: a value per hour times the duration, multiplied by weightings. You set those for task priority, equipment criticality and service contract type.

Against that sit a cost per hour, a cost per overtime hour and a maximum of travel hours per technician type. A scheduling SLA type with its own curve sets how quickly a job becomes urgent. Give a type of work a base value of zero, and the engine will not schedule it at all.

What comes first is therefore a business decision expressed in numbers. Operations and service make it together.

What does the engine do when the day changes?

It keeps calculating and changes only what has not been fixed yet. IFS calls this ‘always optimizing’. If another schedule yields more value after costs, PSO changes the part of the plan not yet assigned. The engine does not move assigned work by itself.

PSO sends a first plan to IFS Cloud once it reaches a quality threshold you set, or after a maximum wait. Changes follow, by default no more than once a minute. The planner sees them in the Dispatch Console, on a live Gantt chart.

Technicians feed the rescheduling too. In IFS Cloud, their expected arrival time goes to the planner, and the engine adjusts the job’s start to match.

According to Erry, AI also brings clear communication to everyone involved. Think of a customer update (‘Henk will be there in an hour’) or a team instruction (‘Call customer X with this message’). Turning arrival times into a promise you can keep is covered in our article on service planning and SLAs.

What does the planner still decide?

Everything where a customer promise or an exception outweighs the maths. The engine proposes. People decide:

• Which values, weightings and costs the engine uses.
• Whether a suggested allocation goes ahead. In the Dispatch Console, the planner accepts the engine’s suggestion or changes it.
• Which work is assigned or unassigned automatically. The Schedule Dispatch Service only does this under commit rules that you define.
• How to handle an exception, such as a customer who only wants Henk.

The planner’s job shifts from juggling and phoning to looking after the rules. Knowledge now in one planner’s head moves into the system. Eqeep sets up those planning rules and priorities together with your planners.

Can you test what a change would mean before you make it?

Yes, but not with the engine that plans the next few days. IFS’s 2021 PSO brochure describes two modules for longer-term questions, IFS WISE and IFS Advanced Resource Planner (ARP).

With WISE (What If Scenario Explorer), you test an idea before you commit to it. You work out how many technicians you need, with which skills and in which locations, and the KPIs they can then achieve. That shows what a large new contract or redrawn territories would require.

ARP builds rosters that respect working time rules, shift patterns, skills and breaks. Those rosters can then go to the engine. Download the brochure for an overview of the PSO modules.

What data does the engine need?

The skills, availability and location of your people, job durations, travel times, your stock levels and your customer agreements. Our article ‘How real can artificial be?’ explains why data is the foundation of any AI application.

A skills matrix is more than a list of names. In IFS Cloud, a resource group can carry a proficiency that makes a job shorter or longer. Certificates must match what the task requires, and so must competencies if the scheduling configuration includes them.

Realistic durations come from completed work. With the Adjust Duration option, IFS uses the average duration of completed jobs of the same type at the same location.

PSO calculates travel time on real road networks, so the engine needs valid map coordinates. Shifts and breaks only go across once they have been generated.

How do you start planning smarter?

With a pilot of limited scope. Know your team, your resources and your SLAs, because only then can AI do its work. A sensible order:

1. Establish a baseline for travel time and cost per job before you set targets.
2. Choose one service organisation or site. In IFS Cloud, you connect a scheduling dataset to it.
3. Let the engine make suggestions first, and have your planners do the assigning.
4. Switch on automatic assignment only once the suggestions hold up, and compare the result with your baseline.

If your team is small and your daily schedule rarely changes, an optimisation engine adds little. IFS Cloud also supports manual scheduling in the Dispatch Console, which may be enough.

In the next blog, Erry looks at AI in sales and marketing. The closing blog covers how to become AI-ready. Where would a scheduling engine save your planners the most work? Start with an analysis.

Questions about AI scheduling in operations

Does the engine consider a preferred technician?

Yes. According to the IFS documentation, PSO schedules work by skills, cost, the preferred technician for the job, location, value and urgency. If a customer accepts only that technician, the planner decides.

What is the difference between route planning and scheduling optimisation?

Route planning finds the shortest route past jobs already shared out. Scheduling optimisation also decides who does which job and when, based on skills, availability, location, cost, the value of the job and the SLA deadline.

Does the engine move work that is already assigned?

Not by itself. PSO changes only the part of the plan not yet assigned. Assigned work changes when the planner moves or unassigns it in the Dispatch Console, or under commit rules you define.

Where does your schedule get stuck?

Tell us where you stand with IFS. After one conversation you will have a clear next step. Bring a day that did not go to plan. Discuss your IFS question.