Intro

Whether you’re building a scheduling solution for employee rosters, field services, or machines, it should be visually intuitive and smart enough to handle constraints, preferences, and changes. Don't settle for building another simple planning tool when you can implement a system that optimizes plans, does more with the same resources, and adapts to shifting scenarios.

This is exactly what we help you achieve with our Timefold x Bryntum partnership.

Timefold is an AI-driven constraint solver for scheduling, shift planning, and routing. Bryntum Scheduler is a high-performance, drag-and-drop timeline UI component that renders interactive schedules. Together, they allow you to build optimized and attractive planners.

How it works: A technical overview

Timefold and Bryntum are both modular and technology-agnostic. You can plug them into any system architecture with minimal friction, whether it's a monolithic, microservices, or hybrid setup.

Timefold

Timefold provides REST APIs that solve complex planning problems such as shift and job scheduling, enabling developers to offload constraint-heavy logic to a dedicated optimization engine. Once a scheduling request is submitted, Timefold returns an optimized solution in JSON format, making it easy to integrate with any system.

Bryntum Scheduler

Bryntum Scheduler is a powerful UI component for intuitive task and resource management.

Fully framework-agnostic, Bryntum Scheduler slots easily into any frontend stack (React, Angular, Vue, or plain JavaScript), allowing developers to build responsive, user-friendly scheduling interfaces with minimal effort.

Integration flow

Your backend system serves as the single source of truth, so that you retain full ownership and control of your scheduler. Timefold and Bryntum are powerful tools for visualization and optimization, but they are not intended to function as data stores. Supplying Timefold and Bryntum with the necessary scheduling data from your own system ensures consistency, prevents duplication, and simplifies data management.

Display data in the Bryntum Scheduler

To display a schedule to your users, the necessary data must be available in your browser.

Bryntum components support various data-loading strategies. You can use the Crud Manager or load data directly into a specific store via a URL. The Crud Manager is especially handy for loading and saving data to and from your backend. You can also integrate WebSocket connections to enable real-time data exchanges between Bryntum components and your source-of-truth system.

Bryntum manages your data using stores, which are reactive, in-memory data repositories that:

• Load and hold scheduling data (in the form of tasks, resources, and assignments)
• Support sorting, filtering, syncing, and all CRUD operations
• Bind automatically to the UI, so that any changes to the stores are instantly reflected in the interface

Note: These stores are not related to the browser's local storage. Any unsaved or unsynchronized data will be lost when the browser is closed.

The data loaded by Bryntum Scheduler doesn’t need to be a complete plan. You can add unplanned events to the UI, so users can plan them manually with Bryntum's drag-and-drop functionality.

Optimize plans with the Timefold APIs

Available as JSON REST APIs, Timefold’s optimization models do the hard planning work for you. No matter the optimization model you use (employee shift scheduling, vehicle routing, or job scheduling), all Timefold APIs follow a similar integration pattern:

1. Gather all relevant planning data: The data you need depends on your chosen optimization model and the constraints you want to enable. Timefold’s APIs only enable certain constraints if you provide the relevant data. The more data points you provide, the more constraints will be active. For example, when scheduling employees, employee costs are only taken into account when you provide the costGroup of the employees.
2. Send that data to the Timefold API to start an optimization run: The Timefold Platform then optimizes the schedule so that it complies with all activated constraints, such as employee availability and skill matching.
3. Monitor the progress of the optimization run by calling the Timefold API to retrieve its status: A single run can take anything from a few seconds to several hours, depending on the scale of the planning problem. The Timefold API always returns the current best solution found so far, along with the current status of the run.

You can implement long polling to continuously fetch updates until the optimization is complete. Alternatively, you can configure a webhook that is triggered when the optimization run finishes.

Customize your integration

Timefold and Bryntum complement each other perfectly, and you can choose how far you want to integrate with either of them. Consider a basic example of backend integration:

• A user clicks Optimize in the Bryntum Scheduler UI.
• The backend collects relevant scheduling data (such as tasks, resources, or constraints), which it then sends to Timefold.
• Timefold processes the input and returns an optimized schedule.
• The backend stores the result and updates the Bryntum Scheduler with the new data.
• The user reviews the updated schedule in Bryntum, then manually adjusts the plan and persists their changes as needed.

Users also benefit from the following optional features:

• Manual edits and score analyses: Use Timefold’s scoring API to analyze the impact of manual changes made in the Bryntum UI.

• AI-assisted recommendations: Let Timefold suggest the best match for a task, rather than leaving it up to manual guesswork.

Alternative: Frontend-only integration

An alternative approach is to call Timefold directly from the Bryntum UI. While this may at first seem easier for existing Bryntum users, it comes with some major downsides:

• The UI needs to gather all the data for the Timefold API, which is not always feasible. Sometimes, the dataset is too large to load in one go, or the data required for planning is not exposed to the frontend.
• Optimizations for large-scale planning problems can run for multiple minutes or even hours. Because your users won't be looking at the Bryntum UI for all that time, you need somewhere to store the component state.
• You would still require a proxy to check your application authorization and to hide the Timefold API keys from your end users.

Note: Frontend-only integration may be okay for demos or prototyping, but it is insufficient for robust systems.

Why trust matters

I love being a dad. It’s really amazing to see this little human learn and explore every day with that child-like curiosity. Kids are so full of “why” questions.

• "Why don’t fish drown?”
• “Why can’t I play on my video game console today?"
• “Why can’t I have ice cream for breakfast if it's made from milk?”

Sometimes we take our time to answer. Other times when parents are stressed, juggling four others things at the same time or are just plain tired, they sometimes whip out the universal catch-all answer: “Because I said so!”.

I don’t like that answer. It’s a shortcut. It ends the conversation. Worst of all, it shuts down curiosity and that is something we can’t afford to lose, especially in this age of ever-more complex systems. Yet, this is also what a lot of scheduling software does.

An algorithm spits out a schedule and the experts, human planners are left with questions:

• Why did Alina get her preferred shift but Myey didn’t?
• Why not fix Thomas’ electricity problem first before moving into the city for the other assignments?
• Why assign Pieter when Maarten lives closer?

When the only answer to those questions is “because the algorithm said so”, planners will not trust the solution. And without trust, they’ll toss the plans aside and will stick to the techniques they have been using before. Our solutions do not just need to find the best schedules… they need to be able to explain them as well.

From explainability to trust

If we want people to accept the output of our planning systems we should give them a system worthy of their trust. A big part in building a trustworthy system is being able to answer the questions mentioned above. It helps you move from blind trust to informed confidence.

In our experience building Timefold, explainability leads to 3 major benefits:

1. Clarity

Giving planners answers to all their “why” brings clarity. They are able to understand why the schedule is what it is. They gain deep insights and are not forced to blindly accept a plan.

With our Timefold solutions, we track constraint violations, how much they affect the outcome, and which decisions caused them. This transparency helps planners understand the trade-offs. At a higher level we also transform these constraints into clear-cut KPIs, making it easier to reason about them.

2. Error backtracking

If something breaks in the real world, you need to know why. In optimization systems which operate like black boxes, it can be hard to figure out exactly where things went wrong.

We have captured this in our Score Analysis functionality which allows you to analyze any schedule, even when it’s not been created by Timefold. This gives planners a powerful tool to diagnose issues and avoid future errors.

3. Insightful adjustments

Planners often make tweaks based on gut instinct. They want to make changes and see the impact of those changes on the schedule.

Next to the Score Analysis mentioned above, Timefold allows you to compare two plans. If new work needs to be assigned to a resource Timefold’s Recommendations assist the planners in making an informed choice.

4. Strategic insights

Explainability in planning systems benefits more than just end users. For decision-makers, they reveal actionable patterns that support strategic decision-making at the executive level.

We're building features in our platform that display operational imbalances and indicate room for improvement.

Clarity, Error Backtracking, Insightful adjustments, and Strategic insights all sprout from the core concept of explainability, and contribute to the trustworthiness of a system. Planners stop overriding schedules, collaborate with the planning tool and are able to solve larger problems with ease, while leaders can use the planning for better decision-making.

In short, they stop seeing PlanningAI as a threat and start seeing it as a planning partner.

Don’t let "Because I said so" kill your plans

We wouldn’t accept that answer from a parent. We shouldn’t expect people to accept it from a planning engine either. In fact, taking a moment to focus on explainability is just as important as finding the “best” possible schedule. A great schedule isn’t just optimal, it’s understandable.

So next time someone asks, “Why this schedule?”, make sure your system can answer with something better than “Because the algorithm said so?!?”.

Some things just belong together.

Timefold’s PlanningAI engine optimizes complex, large-scale scheduling problems in real time. Bryntum visualizes these into beautiful, interactive schedules that are ready to use. Put them together and you finally have the intelligent scheduling solution you always needed.

Why this partnership matters

• See the magic: When optimization becomes visible, it earns trust. Clear, interactive schedules show exactly what changed and why.
• Adopt at speed: Visualization democratizes the complexity of optimization, accelerating buy‑in across your organization.
• Deliver real‑world impact: Lower costs, a greener footprint and higher service levels, now with proof your stakeholders can watch unfold.

A mission, super‑charged

From day one, we set out to free the world from wasteful scheduling. Enterprises across the globe already rely on PlanningAI to slash overtime, cut driving time and delight customers. Yet the biggest barrier to scaling optimization has always been perception: “It’s too complex.” That ends here. With Bryntum’s polished components seamlessly wired to Timefold’s APIs, complexity becomes clarity.

Embed once, optimize forever

Whether you're managing a workforce, running logistics, or routing field teams, our optimization models drop straight into your system with a single API call. And if you add Bryntum, you get fast, clean and responsive Gantt charts, resource calendars, and task dependencies. While our engine handles hundreds of constraints, thousands of tasks, and real-time changes effortlessly, Bryntum brings the clarity.

For the real world

It’s time to pull planning optimization out of the specialists’ corner and put it in the hands of every team that needs it. Together, Timefold and Bryntum are turning algorithmic power into everyday productivity.

Want to learn more? Contact Bryntum, or talk to us!

Stuck in the past

Despite the billions poured into digital transformation, most companies still schedule the same way they did in the '90s. I’ve seen manufacturers, governments, hospitals,… with mission-critical operations still relying on Excel. Or worse: paper.

Really, I am not exaggerating. I’ve witnessed it firsthand multiple times. When they showed me their production scheduling, they:

• Handed me a stack of paper.
• Or showed me several Excel files, none of which actually visualize the schedule in a Gantt or timeline chart.
• Or introduced me to their chief planner, who spends 12 hours a day manually moving around post-its on a giant piece of paper until a feasible schedule magically appeared. This was the prime reason they couldn’t scale operations.

Madness.

And every time it gets to me: Automation is not a given, let alone optimization. Not even close. We have the technology to explore space, while some still haven’t reached the Cape of Good Hope. But in order to catch up, we need to understand the complete timeline.

I broke it down into 4 stages.

Level 1: Freeform manual scheduling

Planning on paper, or in spreadsheets. No guardrails, no validation. If someone assigns two tasks to the same person at the same time, or exceeds the capacity of a machine, it’s all good. Until it blows up in execution. Your schedule is only feasible because of  the human planner behind it. When that planner leaves or goes on vacation? Chaos. Retirement? Disaster.

Level 2: Verified manual scheduling

This is the drag-and-drop utopia many ERP vendors promise: the planner does the thinking, and the software nods politely. It might scream when hard constraints are violated (“You can’t schedule John for two shifts at once!”), but it never lifts a finger to help assign the shifts itself. Your schedule is still only as efficient as the human planner behind it. And when they are absent, production efficiency noticeably decreases.

Level 3: Automated scheduling

This is the first real leap. Press a button, and the software assigns work. Sounds perfect until you realize it’s just a glorified greedy algorithm, assigning tasks one by one in isolation. It’s still only as good as your human planner. Or even worse. There’s no understanding of the true complexity. You’re still leaving a lot of money on the table.

Level 4: Optimized scheduling

Now we’re talking. This is where PlanningAI steps in. Instead of myopically assigning tasks one at a time, it takes a holistic approach. It juggles your hard constraints (skills, time windows, overtime limits) and your soft goals (fairness, fuel costs, customer satisfaction) to deliver a globally optimized schedule. All with one button.

And even then, real-world planning is messy. Some schedules still benefit from manual overrides. Or goal tweaking. That’s why our PlanningAI platform supports mixing and matching verification, automation, and optimization seamlessly. Because planning is not a one-size-fits-all operation. It’s a living, breathing process that needs to adapt in real time.

If you feel like it's time to level up, let's talk.

Over the past few weeks, I’ve had a lot of great conversations with developers, planners, and other professionals at various conferences. Usually, they’re pretty impressed by what PlanningAI (and Operations Research in general) can do and has been doing for decades. However, there is one particular question which always bubbles up: “What if something happens and the plan no longer works?”

Excellent question!

One of my university professors imprinted the following quote into my brain: “Change is the only constant”, which is a quote often attributed to the Greek philosopher Heraclitus, who figured it out way back around 500 BC. Yes, dealing with change has been ruining carefully crafted plans since humans first tried organizing anything.

In the realm of planning and optimization, change can present itself in different ways: An employee calls in sick, a truck decides today is a great day for a breakdown, or a customer ghosts their appointment. Suddenly that beautifully optimized solution becomes irrelevant, sometimes before you’ve even finished your morning coffee.

We are forced to face reality: the real world doesn’t care about our perfectly optimized schedules. Or as another philosopher and occasional boxer put it: “Everyone has a plan 'til they get punched in the mouth” - Mike Tyson

Rolling with the punches

The quote “change is the only constant” isn’t just a sad pointing out of facts. It basically calls upon us to stay flexible in our plans and adapt to new circumstances, because change will come.

That sounds nice doesn’t it… “just adapt to change”. It’s a bit more involved than that.

The first thing we should do is an impact analysis: “What does a certain change mean for our business”? In the case of an employee getting sick, we should ask ourselves, “How critical is this employee?” and “What if we do not replace them?”

The simplest option is to not solve the problem at all, but that often isn’t a feasible path forward. Usually, the plan needs to be adapted, which leads us to the next question.

How much can the plan change?

When a plan is already in motion, abrupt changes can cause chaos. For example, if field technicians are out solving issues, you can't just send them racing off to handle emergencies every hour.

One approach to deal with this is to freeze a part of the plan and let a solver re-optimize the schedule. This will make sure that as much of the plan as possible stays the same. In the field technicians case, we might freeze any assignments for the next hour.

Freezing assignments isn’t always the best approach as it’s sometimes difficult to determine what should and should not be ignored when re-planning. Instead, we could use a Minimize Disruption Constraint. This constraint will compare any newly generated solution with the previous active plan and penalize that solution for each change. This approach nudges the solver to keep as much of the original plan intact as possible.

These disruption constraints can be flexible, penalizing immediate schedule changes more heavily than those further out. That means you can smoothly adapt schedules while keeping your team happy.

These two mechanisms, Freezing and Minimize Disruption Constraints, could also be used together, where you might freeze the next hour as you absolutely want no abrupt changes but do accept some disruption afterwards. This combines minimal immediate chaos and manageable long-term adjustments, the best of both worlds.

Communicating changes

Adapting the schedule is only the start. You still need to communicate the changes clearly to everybody involved. If you have on-call staff, the phone-call to come to work shouldn't be a surprise. If you need to call upon someone who normally wouldn’t be working, that’s another thing.

In these situations, 2 things are fundamentally different from regular planning.

1. If people are unexpectedly going to be asked to work, you should probably call them first if they are ok with that before planning them in. They might be out of town or otherwise engaged.
2. In these situations, achieving an optimal plan becomes secondary to maintaining clear communication and avoiding chaotic changes.

In Timefold Solver this use case is a first-class citizen. With the Assignment Recommendation capability, Timefold Solver combines a greedy algorithm with incremental score calculation to swiftly identify the best available options for fitting the change in the existing schedule.

The recommendations also give some insight into the impact on the overall schedule, so a manager can make an informed decision on who to call into work. This human-in-the-loop approach brings together the power of AI with human intuition which is essential for successful planning.

After the crisis has been resolved the new plan could be re-optimized using the freezing and disruption minimizing constraints.

Embracing change

The move to real-time planning isn’t just technical. It’s philosophical. It means embracing change not as a problem to be avoided, but as a constant to be managed.

In a way, this is the real power of PlanningAI. Not just building smart plans but building resilient plans that stay smart, even when the world decides to punch you in the mouth.

Let’s make plans that adapt.

Operations Research is, and has always been, about solving one of the world’s toughest challenges: planning problems. Finding the best way to allocate resources, schedule shifts, or route deliveries under an abundance of real-world constraints. It is widely accepted that optimizing those problems will make the world a better place, the potential value has never been in question. But the road to real-world adoption has been rough. With PlanningAI, we can alter the narrative.

[highlight]

What is PlanningAI?

PlanningAI is a reckoning for Operations Research. A much needed rejuvenation of the OR space driven by refocusing efforts towards solving real-world complexity at scale, fulfilling all business requirements. It’s a type of Artificial Intelligence that automates and optimizes complex planning, scheduling and routing problems for production use.

[highlight]

Why we believe PlanningAI is a blessing for Operations Research

The unfulfilled promise of OR

For decades, applying OR in practice has been slow, expensive, and risky. It requires niche expertise, complex code, and months to deliver a good POC, only to require some more months to integrate, customize and scale. Additionally, there’s onlly a small pool of hyper-specialized experts that can actually optimize complex planning problems.

Is it because of its sheer complexity, poor marketing, or lack of perceived value that a lot of businesses don’t even think about starting OR projects? We can only make assumptions.

Time to value

However, clarity on and proof of value has always been one of the obvious pain points of OR. PlanningAI solves that. In the first place by translating optimization gains into business KPIs, and in the second place by proving value early. With standardized models that factor in 95% of common constraints, the time-to-value PlanningAI offers is unprecedented. Integration becomes a glue-code job, not a multi-month POC/build/integration saga.

That’s what makes PlanningAI such a shift. It takes the deep rigor of Operations Research and makes it practical, scalable, and ready to plug into real operations.

Risk reduction

As mentioned above, traditional OR projects are long, complex, and expensive. On top of that, the biggest risks show up late. Businesses typically only find out near the end whether the model can actually handle the full set of constraints at scale. By then, you’ve already burned months of budget and resources, often amounting to millions of dollars.

PlanningAI flips that risk profile. Because it starts with prebuilt models that cover the common constraints, implementation is shorter and more predictable. Customization happens quickly, in weeks, so the technical risk is front-loaded. That means less uncertainty, faster validation, and a far lower chance of failure down the line.

Complexity without compromise

The time-to-value and risk-reduction don’t compromise the ability to handle complexity. On the contrary. PlanningAI doesn’t dilute the science but operationalizes it. It works across domains and scales with problem size. Whether you’re planning 50 or 50,000 deliveries, it handles complexity without compromise.

Replanning in real-time

It brings automation where there used to be manual work, optimization where there used to be automation, and speed where there used to be delays. Planners can now generate optimized schedules in minutes, test different strategies, and adapt on the fly. Replanning becomes real-time, not rework.

Results, not in a distant future

Reduced costs, better resource utilization, lower emissions, and higher employee satisfaction… The results from a PlanningAI project are the same as from traditional OR projects. The difference is their accessibility. By using predefined constraints, enterprises can tweak their solution to reach their business KPIs. And importantly, gains show up across industries, from healthcare and logistics to manufacturing and government.

Planners empowered

And it’s more than better outcomes, it’s also about better tools for planners, dispatchers, or other people in charge of allocating resources. Human planners will gain superpowers. With PlanningAI, they can test ideas faster, adapt to last-minute changes, and deliver better plans in a fraction of the time. In some cases companies save hundreds of millions per year.

Democratization of OR

The results of PlanningAI are accessible, and so is the technology behind it. What used to be locked away in cryptic equations and niche academic papers is now packaged in standardized, production-ready models. PlanningAI translates optimization logic into readable, maintainable code, and makes it understandable in thorough documentation. That means teams no longer need a PhD in constraint programming to build or adapt a planning solution. Any developer can understand the model, tweak it, and deploy it. This shift democratizes OR. It opens the door to a much broader group of builders, even onboarding  and support are straightforward.

Mature technology for production use

Unit tested, peer-reviewed, versioned, and production-ready. PlanningAI integrates cleanly with existing systems through a standard REST API, so no need for custom middleware or black-box magic. It fits into your architecture, not the other way around. Without the months-long consulting projects, enterprises go from idea to implementation fast. No brittle glue-code. Just powerful optimization, standardized and ready to plug in.

PlanningAI makes Operations Research operational

In short: PlanningAI delivers what OR always promised. It takes the crown jewels of academic optimization and puts them in the hands of the people running operations every day.

Origin

The term "Planning AI" emerges naturally from the versatile root word "planning," originally from the Old French plan meaning "ground plan" or "map." Historically, planning implied detailed thinking ahead, encompassing both short-term logistics and long-term strategy.

Over time, especially in multilingual and global business contexts, "planning" has expanded to naturally cover various operational scopes, from immediate scheduling and routing to broader strategic foresight.

Practical

In practice, modern organizational needs blur distinctions between short-term schedules, operational logistics (routing), and strategic decision-making. Consider scheduling: it directly impacts daily efficiency, customer service quality, and staff morale. Routing is a close sibling to scheduling. It integrates geographic constraints, logistical decisions, and real-time adaptability. Together, scheduling and routing form the tactical layer of organizational efficiency.

Strategy

At the strategic layer, PlanningAI helps give insight into larger-scale decision-making challenges, such as resource investment, market entry timing, and capacity planning. The strategic horizon defines parameters within which tactical operations like scheduling and routing are optimized.

Critically, restricting "PlanningAI" solely to long-term strategy ignores this practical, linguistic, and historical reality. Its inherent versatility makes it an intuitive, logical term to encapsulate a spectrum of intelligent decision-making processes.

Intention

Therefore, PlanningAI, as a term, intentionally captures scheduling, routing, and strategy because businesses naturally function across multiple timeframes and complexities.

Recognizing this breadth enables organizations to leverage comprehensive AI-driven solutions for greater agility, foresight, and sustained competitive advantage. Ultimately, embracing Planning AI in its broader meaning aligns closely with how organizations practically operate and how language organically evolves to reflect real-world use cases.

In other words: We believe PlanningAI is the world's best solution to complex, large-scale planning problems.

It seems like you can’t scroll through a tech feed these days without hearing about Generative AI. From ChatGPT writing blog posts, to AI-generated art taking over Instagram, GenAI is everywhere. And just to be clear, we’re hardcore users ourselves. But there’s a misconception we want to clarify. Not every AI falls under the GenAI banner. Our DevRel Tom Cools already wrote a compelling blog post on the different kinds of AI, and how each kind has different capabilities. However, in this blog post, we’ll explain 6 major differences between PlanningAI and GenAI. Both powerful, both AI, but fundamentally different.

1. Black box vs. explainable

• Generative AI often relies on deep neural networks that are incredibly complex. The model’s inner workings can be difficult to interpret, hence the “black box” label. Even developers can struggle to explain exactly how the system arrived at a certain output.
• Planning AI, on the other hand, uses methods that are more transparent. It’s built on well-established algorithms (like constraint programming, linear optimization, or rule-based systems) that are easier to trace. Users can see how decisions are made, step by step, which is especially important in industries where compliance and explainability are paramount.

[highlight]

In many business applications, stakeholders demand to know why a decision was made. Planning AI delivers that explainability. Generative AI is incredible for creative or less regulated contexts, but in scenarios requiring a clear rationale (think supply chain, workforce scheduling, or compliance-heavy fields), an explainable system wins every time.

[highlight]

2. Non-deterministic vs. deterministic

• Generative AI is non-deterministic. That means it can produce different outputs even when given the same prompt. This variability can be great for creative applications, but it can also lead to inconsistencies.
• Planning AI is deterministic. Give it the same inputs, and you’ll get the same outcome every single time. This consistency is crucial for businesses that rely on predictable results, like production scheduling, logistics planning, or financial forecasting.

[highlight]

Predictability is often the key to operational success. When you need reliable, repeatable results, deterministic models are more appropriate. If you’re aiming for creative inspiration or novel outputs, non-deterministic models shine.

[highlight]

3. New tech vs. proven tech

• Generative AI feels like the “new kid on the block.” While deep learning has been around for a while, the explosion of large language models and image generators is relatively recent. This newness excites people but also means there’s still a lot to figure out: best practices, regulations, ethical frameworks, etc.
• Planning AI has been around for decades. Methods like optimization, constraint programming, and rule-based systems are proven in countless real-world scenarios. They’ve been tested and refined over time, offering reliability that’s hard to beat.

[highlight]

Mature, proven technologies often come with extensive documentation, robust user communities, and well-established integration paths. If your business requires stability and a track record of success, Planning AI can offer peace of mind.

[highlight]

4. Generating new content vs. optimizing what exists

• Generative AI is about creating. Be it text, images, or even code snippets. It’s an amazing fit for brainstorming, content creation, and personalized marketing materials.
• Planning AI focuses on optimizing existing resources and processes. Instead of inventing a new image or paragraph of text, Planning AI schedules your workforce more efficiently, optimizes complex routing problems and helps allocate resources across multiple projects.

[highlight]

If your biggest challenge is to streamline operations, cut costs, or improve your workflow, Planning AI is designed for those optimization tasks. Generative AI can create fresh ideas, but it might not be the best fit for strictly operational goals that need a laser focus on efficiency.

[highlight]

5. Large training datasets vs. manually programmed algorithms

• Generative AI: Creating an LLM requires massive amounts of training data. These large datasets consume enormous computational power (and, yes, electricity) to reach impressive accuracy and capabilities.
• Planning AI uses optimization algorithms that are preprogrammed by experts. While some data may be involved, it doesn’t require training. Instead, Planning AI relies on logical constraints, heuristics, and well-defined optimization strategies.

[highlight]

Collecting, cleaning, and maintaining huge datasets can be expensive and time-consuming. If your business doesn’t have that kind of data. Planning AI can be far more practical to implement and maintain.

[highlight]

6. Environmental impact: “Burns” vs. “Saves”

• Generative AI is computationally heavy. Training large models can consume tons of energy, which raises environmental concerns.
• Planning AI requires less compute power. It’s more about applying targeted, efficient algorithms rather than brute-forcing through billions of data points. As a result, it tends to have a smaller carbon footprint. Additionally, planning AI is used to optimize issues like vehicle routing, which has a massive impact on CO² emissions.

[highlight]

Sustainability is becoming a core metric for many organizations. While it’s not always a direct apples-to-apples comparison, if your company values a lower environmental impact, Planning AI is often more resource-friendly.

[highlight]

Where does Timefold fit?

We’ll be the first to admit we’re big fans of Generative AI, some of us are even using it to help write code or jump-start content. However, our product is built on PlanningAI principles. It’s explainable, deterministic, and designed to optimize existing processes rather than create new content. This makes it ideal for businesses that need reliable, repeatable, profitable, and environmentally conscious solutions.

Our takeaways

• Generative AI and Planning AI are both “AI,” but they solve very different types of problems.
• Generative AI shines at creating new content and inspiring innovation, whereas Planning AI excels at optimizing complext processes at scale and delivering predictable results.
• Both have unique strengths, but if you need transparency, consistency, and a smaller environmental footprint, Planning AI may be the better fit.

As a cat person, I’m very happy the internet has embraced cats as the “ultimate feel-good” subject when it comes to pictures and videos. And now with LLMs and GenAI the fun is endless! Want a cat sitting on a crocodile while playing the banjo? Boom!

Generated by ChatGPT

Or a cat in a tiny top hat hosting a stand-up comedy show for a crowd of impressed goldfish, here you have it!

Generated by ChatGPT

Of course this is purely for entertainment. LLMs have proven themselves to be very valuable tools that help us brainstorm, proof-read e-mails or even help us rewrite texts for specific audiences. The progress in the past few months has been mind-blowing, which might suggest they're the answer to everything. At least, if you believe the hype machine.

But when it comes to solving practical, real-world problems, they’re not there yet.

LLMs and everyday planning problems

My neighbor Jamie works at a hospital and creates the work schedule for the nurses and doctors. It’s a logical nightmare juggling availabilities, contracts, and other employee preferences. Every week, she spends hours creating the schedule, but there is always somebody who thinks they got the short end of the stick.

Our local plumber, Robert, is always in high demand, constantly driving around the neighborhood. His services are in high demand, but he wastes valuable time on the road because he isn't tackling jobs in the most efficient order. Smarter scheduling would mean helping more customers and spending less time driving.

With all the power that people subscribe to tools like ChatGPT, creating such schedules should be a walk in the park? Right?

Reality is however disappointing: LLMs can’t deal with these tasks very well. Most of it comes down to the sheer size of the problem space: the total number of possible combinations of input parameters. Even a seemingly simple problem like assigning shifts to 10 Employees for a week has a problem size of 10⁶³ combinations. That takes a lot of processing power to solve even with specialized algorithms, let alone trying it with a general language model.

7 days, 3 shifts per day, on average 3 employees per shift, yes that’s a 1 with 63 zero’s behind it!

Sometimes LLM models do find a feasible solution. This is especially true for the newer reasoning models. Those seem to think through the problem and can create a feasible schedule on very small datasets. As these models can reason about discrete steps, they break down the creation of a schedule to smaller steps, just like a human would.

Recent updates have given LLMs to replicate human reasoning.

When you try medium sized datasets, LLMs run both into the problem space issue and in some cases even context window limitations due to the amount of tokens in the input. Although we do believe the latter is a temporary problem as the technology evolves (looking for example at Gemini with a context window of 2 Million Tokens!), the former is going to be much harder to overcome… at least without some help.

A solved problem, almost forgotten

By their very nature, LLMs just predict the next token and “the next token” isn’t what we are looking for when trying to solve planning problems. Luckily, for most gaps in current LLM capabilities, we can look back at “older” forms of AI which resolve these gaps.

The different layers of AI, Timefold fits into the outer layer of old-school AI.

When looking at our problem, we have to peel back the layers of AI evolutions. As mentioned, LLMs and GenAI aren’t the solution here.

Deep Learning and Machine Learning are excellent at pattern recognition and learning from data. Unfortunately, the learning part is often the problem.

ML solutions require large volumes of high quality data to learn from. Real world planning problems usually have their own unique nuances and are often still being done manually… so there is no large amount of high quality data available. No training data, no training, no Machine Learning model.

We need to go even further back, to the origins of AI. In the “old-school” AI layer, we find a pretty unassuming part of AI called Mathematical Optimization. This blob of techniques has been a proven way to solve planning problems for the past 30 years. Here you’ll find techniques such as Linear programming and our personal favorite at Timefold: Metaheuristics.

These algorithms don’t need a big upfront training set to build a model, as the models are programmed manually by domain experts. They work with enormous datasets and find (near) optimal solutions to your planning problems.

It’s frankly a wonder Metaheuristics and friends are so undervalued and get overshadowed by the commoditization of GenAI. By our estimations, 95% of these planning problems remain unresolved while the answer has been out there for decades.

Maybe it got lost in history? It’s time to rebrand this part of our industry and start our own hype machines. It’s time for #PlanningAI.

[highlight]

What is PlanningAI?

PlanningAI is a type of artificial intelligence designed specifically to handle complex planning and scheduling tasks, and to satisfy the constraints of planning problems. It helps you make better decisions by sorting through countless possibilities to find the best solutions. Solutions that save you time, reduce costs, and improve efficiency.

This form of AI can be a bit unwieldy and overwhelming to operationalize. We realized that and removed as many barriers as possible to make it as simple as it can be. We’ve created pre-built models for common use-cases, ready for consumption through a REST API interface on our Platform.

Using these APIs, we can generate a shift schedule for my neighbor Jamie, or calculate an optimized route for our local plumber Robert, without even needing to know anything about solving these complex planning problems. It’s basically our decade long planning expertise in a box.

[highlight]

Not one or the other, but together

While LLMs can’t handle planning problems efficiently, nobody can deny that they’re an amazing enabler. So instead of staying in our respective corners of the AI-Ecosphere, let’s start being a big happy AI family and leverage each other's strengths for the betterment of all!

PlanningAI can leverage Custom GPT models to help domain experts build better PlanningAI models and help explain them to a larger audience. It can also help transform data into information and translate complex ideas into something even the most technophobic person can understand.

LLMs from their part can leverage the capabilities of PlanningAI through “Tool Calling” functionality or the Model Context Protocol, giving LLMs access to advanced algorithms specialized for solving these kinds of problems.

The fusion of AI technologies is reshaping how we solve planning problems, opening up new and innovative solutions. What a time to be tackling these challenges!