Forum Discussion
Q&A: Frontline
Hi DarrenNash​,
Thank you for taking the time to share such a thoughtful and detailed perspective!
I can understand your concerns and appreciate that you’ve identified areas where you see value in the approach while also asking important questions about how learning expertise and human judgment fit into the process. There’s a lot to unpack here, and I’d like to address your questions as best as I can!
1. What does Articulate mean by “grounded in learning science”?
Articulate’s AI is designed to do more than generate content; it helps structure information into training that supports how people learn. We've engaged with professional instructional designers at each step of our product development over the past 20+ years. This has resulted in professional learning tools that support the application of principles such as sequencing, chunking, scaffolding, retrieval practice, interactivity, and assessment to the formats and experiences our AI produces.
For example, AI-generated training can incorporate practice opportunities, knowledge checks, and quizzes rather than simply presenting information. The prompts and generation workflows have also been refined through testing with our internal learning experts and again with customers.
In addition, creators remain in control throughout: they guide the AI and review, edit, and approve its output. “Grounded in learning science” doesn’t mean the AI replaces instructional-design expertise; it means that expertise is built into the foundation creators work from.
2. Does Frontline analyze a performance need, or does it transform supplied content?
This is an important distinction you’ve made between performance analysis and content transformation.
Frontline is designed to help someone take a training need and supporting source material and turn that into a coordinated set of resources. Nova can use the context and materials provided to help determine how that information might best be structured and presented rather than simply converting a document slide-for-slide into another format.
Frontline shouldn’t be interpreted as a replacement for the broader performance analysis that can happen before an organization or individual decides to build the training.
3. How does Nova establish what learners need to do?
Nova uses the information and source material, as well as input on goals and delivery format that a creator provides to help build a training kit and recommend different ways to present it to learners for practice.
4. How are meaningful scenarios created?
In Frontline, Nova builds a branching scenario from your prompt and source material—complete with dialogue, choices, and feedback tailored to your training needs. Human review remains important here, and you’re welcome to edit the scenario, including the experience setting, which is how your audience will experience what you’re building. For example, a text-only scenario versus a conversation with a character whose expressions change based on the text the user selects.
5. Where does L&D governance sit within Frontline?
One of Frontline's goals is to enable teams outside L&D to create and maintain training while still benefiting from the expertise, resources, and standards established by their organization. In most organizations, L&D needs to play an important role in establishing expectations for what good training looks like, determining when professional review is appropriate, and helping teams apply those standards.
Currently, the creator has publishing permissions and access to the analytics dashboard that shows how the content is performing.
6. How is effectiveness measured?
Measures such as completion, engagement, survey responses, and the questions learners ask can provide useful signals. They can help an organization understand how learners are responding to the training and where questions or uncertainty may arise. Those insights can help teams identify areas to review or improve.
These measures alone don't prove that someone changed their behavior or that the business outcome resulted from the training. That kind of impact may require additional evidence outside the learning, such as other performance measures or observations.
7. How is the AI Tutor governed?
AI Tutor is grounded in the content found in the training. It uses the published content as its primary knowledge source and tailors responses to the specific content your audience is working on. AI Tutor does not browse the web. Responses stay relevant, controlled, and aligned with the author’s intended training experience.
Here’s an FAQ on managing AI Tutor in Frontline that may be helpful!
8. What happens when source material changes?
When source material changes, you can update the content in bulk by chatting with AI to describe the changes you want to make. While AI drafts your changes, you'll see a status message showing how many changes it's preparing across your kit.
Once the changes are ready, AI will provide a quick summary of which artifacts the updates affect. Then, you have the option to Confirm all, Review changes, or Reject all.
You can also save a snapshot within an artifact to roll back to it. Please note that creating a snapshot is a manual process.
9. How are accessibility and localization quality assured?
Thank you for raising this topic. We’re actively working on a Frontline Accessibility Conformance Report (VPAT), which is expected to be available in Q4 of 2026. You can check our Trust Center for updates on its availability.
For localization, it’s best to ensure the source language is correct before translating your kit. You can then review the translations and make updates before sharing the content. You can find more information on translation kits here.
10. How does Articulate see the future role of the learning professional?
I really appreciate you sharing this perspective here. As AI is changing how learning is created across organizations, we see an opportunity for L&D to evolve as well. We've shared more about our perspective in The Acceleration Gap.
Thank you again for taking the time to share your feedback in such detail. This gives us valuable insight into where we can provide more clarity. I’d also be happy to connect you with our team for a more in-depth discussion about Frontline and the questions you’ve raised!
Thank you for taking the time to provide such a detailed response. I particularly appreciate the clear distinction you make between engagement data and evidence of behaviour change or business impact.
However, several of your answers reinforce my central concern.
You confirm that Frontline does not replace the broader performance analysis that should happen before training is created. Nova relies on the creator to supply the need, goals, context and source material. Human review is then required to judge the scenarios, translations and final outputs.
Those are not minor tasks. They contain much of the professional judgement involved in workplace learning.
This creates a significant gap in the proposition. Frontline is marketed to IT, Sales and other business teams so they can create and publish their own training. But who ensures that they have correctly identified the performance problem, that training is the appropriate intervention, and that the source material addresses what people actually need to do?
If the diagnosis, goals or source material are wrong, Nova may simply produce a polished solution to the wrong problem.
We then only hear of it afterwards. This already happens in some cases with Rise modules for example.
It will also be abused by Vendors to create content quickly at a premium rate.
The reassurance that creators remain in control does not entirely resolve this. Human review is only an effective safeguard when the reviewer has the expertise to recognise weaknesses. An SME may be able to verify technical accuracy without being able to evaluate learning design, practice, feedback, accessibility or transfer.
I would also welcome greater clarity around “grounded in learning science.” Sequencing, chunking, scaffolding, retrieval practice and assessment are valid principles, but including these features does not establish that they have been applied appropriately. An ineffective question is still a question, and superficial branching is still branching. Was Frontline evaluated against defined learning or performance outcomes, and will Articulate publish the methodology or findings?
You state that you engaged "professional instructional designers" but that does not establish that those people were experts in adult learning, learning science, performance analysis, cognitive psychology, assessment validity, accessibility or learning transfer.
“Instructional designer” is not a regulated title. It can describe:
- A learning scientist or highly experienced learning professional
- Someone trained in instructional systems design
- An e-learning developer using Storyline or Rise
- A content producer given the title by an employer
Therefore, neither “professional” nor “20+ years” tells us what expertise informed Nova. It tells us that Articulate consulted people holding a particular job title over a long period.
Even if all those instructional designers were genuine learning experts, consultation during product development would still not demonstrate that learning expertise has been successfully encoded into the AI or that Frontline’s outputs improve learning and performance. That requires transparent design criteria and outcome-based validation.
The governance answer concerns me most. You say L&D should establish standards and determine when professional review is appropriate, but you also confirm that the creator currently has publishing permission. This makes L&D involvement advisable rather than necessary.
Standards alone are not governance. Without approval gates, differentiated permissions and required review, Frontline does not necessarily extend L&D expertise across the organisation. It allows the organisation to decide whether that expertise will be consulted at all.
I also appreciate your acknowledgement that completion, engagement, surveys and learner questions do not prove behaviour change or business impact. I hope this distinction will be reflected consistently in claims that Frontline can demonstrate that training “worked” or delivered ROI.
The accessibility position is also difficult. A future VPAT does not mean Frontline is inaccessible, but launching before the conformance report is available prevents organisations from completing a fully informed accessibility assessment before adoption.
Finally, my question about the future role of learning professionals remains unanswered. Saying that L&D has an opportunity to “evolve” does not explain which decisions require learning expertise, when Frontline creators should involve learning professionals, or how the product prevents that expertise from being bypassed.
Learning already appears to be one of the few professional fields in which expertise is routinely treated as optional. Rise reduced the technical barrier to creating online content. Frontline goes further by suggesting that some of the learning expertise now resides in the software.
My concern is not that Frontline genuinely makes learning professionals unnecessary. Your answers demonstrate that analysis and human judgement remain essential. My concern is that Frontline makes it much easier for IT, Sales and other teams to circumvent those processes while believing that “built-in learning science” has replaced them.
Hi DarrenNash,
Thank you for sharing your feedback and concerns so thoughtfully. I’ve shared your perspective with our team, and I truly appreciate the opportunity to connect with customers and have important conversations like this!
Darren, to your final concern, if you take the role of Chief Learning Officer and create a governance document your organisation adopts, you will be the SME empowered to steer teams to best practice with balloon guardrails like in ten-pin bowling. In this process you also ensure the QA testing, release gates, version control etc. are maintained to the level required by the product.
If you check out the Articuland launch there is a few topics Create, Integrate, Orchestrate, and four announcements regarding these topics.
This fits into the Orchestrate and fourth announcement of Frontline.
There is a fellow first guest speaker on stage that speaks of extending his role from being a professional to strategy of L&D for the organisation. He covers these types of concerns you have and the pivot to use the tool to enable and add value for measurable outcomes.
The first thing I've encountered is managing AI spend on a product and ability hasn't existed at last financial year. I'm working on a demo with the sizzle and the governance to be able to inform the budget aspect from teams when that ask arises.
From an outside perspective, I'm looking at participating in this model as a citizen developer of power platform to apply it in the L&D space. The IT department have a governance document ensuring they control everything they are accountable for. I am enabled to prototype and test requirements through pilot solutions, to either inform a specialist to take over the project, or them to approve my spec and grant access to continue an iteration myself. Parallel to how I would govern an SME using Frontline. They may burn through AI credits and hand over to the specialist function, or deliver their product to QA process.
The marketing isn't going to dive into this level of detail, but I would just remember it's only a tool. If the output isn't fit for use (and the intended use will guide the go/no-go before even starting), the evalution and feedback should capture that, by process as any other document type or communication would be judged.
There is no cost in working with the tool to ensure requirements and outcomes meet, but there's best practice SME's will need us to guide them. I've found the workflow Articulate have been doing for a long while with the Betas, getting to first draft and ability to author it as required to final, is very good.
We're now in the era where "Gemini is AI and can make mistakes" is publicly stated in the same page as the query. My enterprise Copilot window has "AI-generated content may be incorrect" ... although OpenAI chasing IPO just say "it is AI" now so can't shake the snake oil habit just yet (they did have a more transparent statement they've now obfuscated by elimination).
Regards,
<- is Leader of the Free World<obf>...(in my head)...</obf>Thanks for the detailed response. I agree that governance, quality assurance and appropriate oversight are essential, but I think we are discussing two fundamentally different things.
My concern is not primarily about governing the production of content. It is about ensuring that the same established, evidence-based principles of learning science and instructional design are applied, regardless of whether the content is created by an experienced learning professional, an SME or AI.
I would particularly challenge the suggestion that an SME should be responsible for providing or defining learning best-practice guidelines. An SME is an expert in their subject matter, not necessarily in how adults learn, how skills develop or how learning transfers into workplace performance. These are distinct areas of expertise. An SME can validate technical accuracy of the material required and identified by analysis but that does not qualify them to determine instructional effectiveness. In my organisation I know very few SME's who are learning experts but we use them to validate and provide the content and material required to answer the question of the problem we are trying to answer and how we create a behavioural change. Based on investigation and further analysis and getting user feedback we determine then the best approach to deliver the content.
Also with Instructional Designers, the title is not a protected one so anyone can take that title. Most ID's I know are not learning experts either, they all have varied roles and responsibilities. Many open up Storyline 360, so they are now an Instructional Designer and can use the title, it does not mean they know anything about learning. When Articulate said they worked with many Instructional Designers, I actually thought, so what? Did they speak to actual learning experts?
Similarly, appointing a Chief Learning Officer or introducing governance documentation does not automatically ensure sound learning design. Governance may control how a product is developed, reviewed and released, but it cannot compensate for a failure to establish whether learning is even the appropriate solution in the first place.
This is precisely why learning needs analysis, performance gap analysis, appropriate instructional strategies, meaningful practice, feedback and evaluation are so important. These are not simply production standards or optional guidelines that can be applied at the end of development.
My concern with Frontline is that it potentially enables people to produce what looks like professionally designed learning without necessarily understanding the principles that make it effective.
I am not questioning whether AI can assist with learning development. Of course it can. I am questioning whether accelerating and democratising content production is being presented as equivalent to applying professional learning expertise.
An SME can establish what is technically correct. A learning professional must establish what, if anything, needs to be learned, why, by whom and how that learning should be achieved.
Those responsibilities are not interchangeable, and AI does not make that distinction any less important.
Hi Darren, you took a significant wrong turn at "the suggestion that an SME should be responsible for providing or defining learning best-practice guidelines." I think you've assumed the role of SME is singular and "other", foreign, not a hat "we" would wear.
To clarify, I suggest if you adopt the role of CLO, then you as the SME for L&D in the org, adopt that top role, apply your expertise with the power of governance to address all your stated concerns. Have the CEO sign to it in a relevant policy update paired with likely IT policy updates around these topics.
Prior to your final concern, much of it seems to be challenging Articulate's claims and description, so I skipped past them as a staff member had replied. If you've been in the game for a few decades, been in the weekly challenges with the community for a couple decades, used the tools, seen the progress, generational architectural shifts v1.0, v2.0, v.... participated in betas, community code block buildathon... I guess the curse of knowledge it's not questions we ask or think about. https://www.articulate.com/about/
Re: what "looks like..." if you jump back to early word processors, marketed to "create professional looking documents". Substitute your concerns for AI or Nova with Spell Check and ID's for Editors - we had then as we have now professionals, templates and style guides, peer review processes. We still use tools for with accountability and process to achieve a target a level of QA. There's just as much democritisation for an intern using a notepad submitted to a typist, but the Editor was accountable for what is published. That
So most of your concerns I read as "doing our job". It seems a foregone conclusion that if you don't do all the fundamental best practices and evaluate they're successful to your outcomes, you may have some consequences before even considering these tools. These also may be more of your problem than the person you see as the problem.
I also think you will feel more comfortable with "democratisation" by acknowledging that "we" do not have much more than formal recognition of what happens in spite of us. Just as "laws of physics" are just our articulation of what actually occurs in nature, "breaking" a law is just breaking our concept of it... nature persists regardless of our perspective. Learning is ubiquitous with human history, and living creatures broadly in their survival.https://www.youtube.com/watch?v=NGaUM_OngaY
So surveying who you fear will bring down this house of cards... in my case other experts are that - professionals or masters of their domain, collectively also with more experience and more learned in my own area than I. Just in Time? Open Source? Agile? MVP? There's so many ways of approaching it.https://www.atlassian.com/agile/manifesto
The 4 Agile values are:- individuals and interactions over processes and tools
- working software over comprehensive documentation
- customer collaboration over contract negotiation, and
- responding to change over following a plan
These values prioritize people, collaboration, and adaptability in software development.
"Sound Learning Design" / not automatic - it's naive to assume anything is automatic! Workflows still require work. A team. Many teams. Effective management of them. Them knowing who is accountable for what, what they're responsible for when, how to communicate when, it goes on and on, and that's just saying what a RACI is. Then there's Support services to be acknowledged so you need to update the abbreviation to make sure you don't neglect infrastructure. Is any of the business even being done yet to get revenue?
In my area, most of what we do gets in the way of what matters, but compliance with law/regulations, legal liabilities, require a hefty burden obstructing the learning, that is mostly practical in-person on the job.Here's a hammer. Here's a sickle. Here's a charka. Articulate are providing tools, how we use them is up to us.
As bloat reducers, if we have some effective new tools that slash the time to first draft to get past the chores... well you can guess what I'm already doing. Hand (food processor lol) churning cream to make butter... (if I ever finish this reply) saved by not copy-pasting as much. It might be nostalgaic for some manual work like milking the cow for tourists. /optimistic-cow
