Key takeaways
"Custom" means different things to different vendors. At minimum, it should mean your organisation's tools, workflows and use cases appear in the training content, not just a logo on a slide deck.
The most common version of "custom" is a standard course with a few swapped examples. That is rebranding, not customisation, and it rarely changes behaviour on the job.
Genuine customisation starts with a needs analysis before a proposal. If a vendor quotes you a price before asking how your teams work, treat that as a warning sign.
Four signals indicate real customisation: role-specific scenarios, your actual tooling in the exercises, content built around your risk and governance context, and measurable outcomes agreed before delivery.
You can test a vendor's claim cheaply and quickly by asking two questions: "Can you show us a sample scenario built for a team like ours?" and "How will we know this worked?"
What does 'custom AI training' actually mean?
Genuine custom AI training starts with your organisation's actual work, not a generic course someone has lightly rebranded. It means the scenarios, tools, roles, and risks in the training content are drawn from how your teams operate, not from a hypothetical company in a vendor's example library.
That distinction matters more than it sounds. A procurement team at a state government agency faces completely different AI questions than a marketing team at a retail group. The tools differ, the governance obligations differ, and the day-to-day workflows that AI is supposed to improve are nothing alike. Training that treats both audiences the same is not custom training. It is a catalogue course with your logo on the cover slide.
The clearest test for real customisation
If you swapped your organisation's name for a competitor's and the training content would still make complete sense, it was never custom to begin with.
In practice, genuine customisation touches at least three things. First, the content itself: the examples, use cases, and exercises reflect your industry, your systems, and your real job functions. Second, the learning design: the depth, pacing, and format match the actual skill level and time constraints of your people. Third, the framing: the training positions AI relative to your specific policies, your existing tools, and the outcomes your organisation is trying to reach.
None of that can happen without a proper discovery process before anyone writes a single slide. A vendor who quotes you a price before asking substantive questions about your workforce, your tools, or your goals has not done the diagnostic work that real customisation requires.
That does not mean every element of a program needs to be built from scratch. A well-structured delivery partner will have a solid foundation to work from. What makes it custom is how intentionally they adapt that foundation to fit you, and how much of your operational reality ends up embedded in the final product.
For a broader look at where custom training fits against off-the-shelf options, the custom vs off-the-shelf AI training article covers the full decision.
What are vendors usually selling when they say 'custom'?
The word covers a wide range. At one end, a vendor swaps your company logo into a slide deck they have run for dozens of other clients. At the other, a team spends weeks inside your organisation before writing a single word of content. Most offerings sit somewhere between those two points, and the distance matters more than most buyers realise.
Here is how the spectrum tends to look in practice.
Logo and colour swap. The course content is entirely generic. Your brand assets appear on the cover slide and possibly in the header. Nothing else changes. Vendors offering this level of "customisation" are essentially reselling a catalogue product with a cosmetic finish. It is the most common thing sold under the custom label.
Scenario replacement. The underlying framework and slides stay intact, but a handful of examples are swapped out. A prompt-writing exercise might reference your industry instead of a generic retail context. This is a meaningful step up from pure rebadging, but the core learning design has not changed, and neither has the depth or sequence.
Module selection. You pick from a menu of pre-built modules and assemble a programme. This gives you control over topics covered and running time, but each module was designed independently, not as part of a coherent journey for your teams. The joins between modules can feel abrupt, and there is usually no thread connecting the content to your actual AI adoption goals.
Facilitation adjustment. The content is fixed, but the facilitator spends time before delivery understanding your context and weaves in relevant references during the session. This depends heavily on the individual facilitator and rarely survives a trainer substitution. It is customisation that lives in someone's head rather than in the programme design.
Genuine bespoke design. The provider conducts a discovery process, maps the programme to specific roles or workflows, builds scenarios and exercises from the ground up, and structures the content as a deliberate learning journey. This is what most buyers picture when they hear the word "custom." It is also what most vendors are not actually offering.
The gap that catches buyers out
The first four approaches are common and often competitively priced. Genuine bespoke design takes more time and costs more, but it is the only version that meaningfully connects training to how your people actually work. Knowing which one you are being quoted on is the most important question to ask before you sign.
None of these options is automatically wrong for every buyer. A logo swap on a high-quality, well-scoped catalogue course can be perfectly adequate if the content genuinely fits your context. The problem arises when a vendor prices and positions a light rebadge as if it were a bespoke build, and a buyer has no way to tell the difference until delivery day.
Which customisation signals are worth paying for?
Not all customisation creates equal value. Some of it costs money without changing what participants actually learn or retain. The signals below point to the kind of tailoring that shifts outcomes rather than just shifting the slide deck.
Scenarios built from your actual workflows. Generic examples about "a retail company" or "a healthcare organisation" ask participants to do translation work before they can apply anything. When a vendor interviews your subject matter experts before design begins, and builds exercises around tasks your teams genuinely perform, that translation work disappears. A procurement officer practising prompt engineering on contract review, rather than on a made-up marketing brief, walks out with something immediately usable.
Role segmentation beyond job title. Splitting participants into "technical" and "non-technical" groups is a starting point, not a solution. Genuine role-based design asks what each group needs to do differently on Monday. A finance analyst using Copilot to summarise board papers needs a different session to a comms manager using it to draft stakeholder updates. If a vendor proposes a single pathway for both, the customisation is surface-level.
Learning objectives tied to your business goals, not the tool's feature list. A vendor who begins the scoping conversation by asking about your tools is starting in the wrong place. One who begins by asking what behaviour you are trying to change, and then works back to the tool, is thinking about outcomes. The difference shows up in how objectives are written: "Participants will understand Copilot features" is a vendor objective; "Participants will reduce first-draft time on weekly reports by using Copilot prompts" is yours.
Assessments that reflect real tasks. Off-the-shelf programs tend to use multiple-choice quizzes because they are easy to build and easy to score. Practical assessments, where participants complete a task they would actually face at work, are harder to design but far more predictive of transfer. Ask the vendor what the assessment looks like before you agree to the scope.
The clearest indicator of genuine customisation
A vendor who cannot describe the discovery process they run before design begins has not built a customisation methodology. They have built a template with a name field at the top.
Manager enablement built into the design. Training rarely sticks without reinforcement from direct managers. A well-designed custom program includes at least a short manager briefing: what participants covered, what they should be practising, and how managers can create space for that practice. If a vendor's proposal does not mention managers at all, ask why.
Delivery format matched to your context. Remote cohorts, mixed-timezone teams, and deskless workers all have different constraints. A vendor who proposes a full-day in-person workshop for a team spread across three states either has not listened or has only one delivery format in their toolkit. This matters more than it sounds, because format affects participation rates, and participation rates affect ROI. See custom vs off-the-shelf AI training for a fuller breakdown of when each approach suits which context.
One practical shortcut: ask the vendor what their standard scoping process looks like and how long it takes. Real customisation requires real discovery. If the answer is a 30-minute call followed by a proposal, the customisation is happening after the contract, not before it.
How should you test a vendor's customisation claim before signing?
The procurement conversation is where vague promises get exposed. Most vendors sound credible in a pitch deck. The gap shows up when you ask specific questions and watch how they respond.
Start with the brief. Tell the vendor you have a particular audience in mind, say a team of finance operations staff who need to use AI to process invoices and flag anomalies, and ask them to walk you through what their discovery process would look like. A vendor who has done this before will ask you follow-up questions: What tools does the team already use? What's the biggest friction point today? How technically confident is the average participant? A vendor selling rebranded content will pivot quickly to showing you a course catalogue.
Ask to see a before-and-after. Request an example of their standard content on a given topic alongside the customised version they delivered for a previous client (with any sensitive material removed). If the two documents look nearly identical, that tells you everything. If the customised version has different scenarios, different terminology, and a different participant profile built into the exercises, that's a genuine signal.
Ask who writes the content. Some vendors outsource content development to freelancers after the sale, with the facilitator only seeing the materials days before delivery. Others have the facilitator involved from the brief stage. Ask directly: who conducts the discovery interview, who writes the scenarios, and are they the same person who delivers the training? The answer reveals how much institutional knowledge actually transfers into the room.
The question that separates real custom from surface-level
Ask a vendor: "Can you show me a version of this course before and after customisation for a previous client?" A confident answer with real evidence is the clearest signal you'll get in a procurement conversation.
Probe the timeline. Genuine customisation takes time. If a vendor promises a fully tailored program in three days from signing, they are either using a very thin personalisation layer or they have a template they're relabelling. Ask how long discovery, content development, and facilitator preparation typically take. Expect something in the range of two to four weeks for a meaningful customisation, depending on scope.
Ask what they will not customise. This one is counterintuitive, but a trustworthy vendor should be able to tell you where their flexibility ends. Some content, particularly foundational concepts or licensed tool demonstrations, may be standardised for good reason. A vendor who claims they can customise absolutely everything, with no constraints, is probably not being honest about their process.
Finally, ask for a pilot or a scoping session before you commit to a full rollout. A vendor confident in their customisation capability will usually offer a discovery workshop or a scoped proof of concept. If they push back on this and want you to commit to the full program upfront, that's worth noting.
The goal is not to catch a vendor out. It's to find one who finds the right questions as natural to ask as you do.
Frequently asked questions
What is custom AI training, exactly?
Custom AI training is a program built around your organisation's actual tools, roles, and workflows rather than a generic curriculum sold to any buyer. At minimum, that means the examples, scenarios, and skill targets reflect what your people actually do. At its best, it also means the structure, depth, and pacing are shaped by a discovery process with your team before a single slide is written.
How much more does custom AI training cost than an off-the-shelf course?
Genuine custom programs typically cost more upfront because of the design and discovery work involved, but the gap is smaller than most L&D leads expect once you account for what off-the-shelf courses exclude. A tailored one-day workshop for a specific team can still sit in a range comparable to a mid-tier public course, particularly when you are training a group rather than individuals. The more relevant question is cost per outcome, not cost per seat. An off-the-shelf program that produces low transfer will cost more in the long run than a well-scoped custom one.
Can't we just ask a vendor to swap in our logo and call it custom?
You can, but it will not produce meaningfully different results from the original program. Logo swaps and minor cosmetic edits do not change the scenarios learners work through, the tools they practise with, or the skill gaps being addressed. Staff will still encounter examples that feel abstract or irrelevant to their role, and transfer to day-to-day work will be limited.
How long does a genuine custom program take to develop?
A well-scoped one-day workshop built from scratch typically requires two to four weeks from initial brief to delivery-ready materials, assuming the vendor conducts a proper discovery session with you early in the process. Longer programs with multiple modules take proportionally more time. Be cautious of vendors who promise a "fully custom" program with a turnaround of two or three days. That timeline is only achievable if the customisation is superficial.
Which roles benefit most from custom AI training?
Any role where the gap between generic AI instruction and real-world application is wide. In practice, that means non-technical teams who will not find standard developer-focused content relevant, specialist functions like legal, finance, or procurement where compliance context matters, and senior leaders who need a program shaped around strategy and risk rather than hands-on tool use. Off-the-shelf content works well for foundational literacy; custom becomes essential once you need people to change how they work in a specific context.
Ready to see what genuine custom AI training looks like?
If a vendor can describe your training programme before asking a single question about your organisation, that is your answer. Genuine custom AI training starts with curiosity about your people, your tools, and the specific workflows where capability gaps are costing you.
Better People builds programmes around what your teams actually do. That means discovery before design, your real systems in the examples, and outcomes you can measure against the business case you made to get the budget.
Want to know what a real scoping conversation looks like?
We'll ask about your tools, your teams, and the outcomes you need. You'll leave with a clear picture of what a programme built for your organisation would actually involve.
