by Craig Casten | Sep 1, 2026 | Articles
Artificial Intelligence (AI) can save awards teams meaningful time, but only when it holds a clear purpose.
In Award Force, AI can help with tasks such as summarising entries, extracting information, identifying themes, drafting feedback, working through calculations and more; it can even do initial moderation, judging or score normalisation—all with humans in the loop. That’s cutting-edge technology wrapped in 15 years’ worth of awards-know-how.
However, the important point is these tasks do not all ask the same thing of AI.
Whether AI is built into an awards platform like Award Force or sits within some external workflow, this principle is useful: AI behaviour should reflect the job it is asked to support.
Award Force enables awards teams to create purpose-built AI configurations using supported large language models (LLMs). Each configuration can have its own name, model, temperature and pre-prompt instructions, giving the AI context about what it is expected to do and how it should respond.
This is materially different from what other awards or AI solutions offer because f in Award Force, you own the flexibility, you control the AI response and you decide where and when to deploy your AI tool. Whatever makes most sense for your awards.
In Award Force, AI is not managing your program without you, it’s not making decisions that should be confined to humans and you are not restricted to what a vendor thinks you might need from AI (or what will make you spend the most AI tokens with them).
Rather than leaving AI to be a vague general assistant that knows little, you are empowered to build a collection of focused tools for different jobs within your awards program.
Take the tour
Despite what the main AI providers want you to believe, the available LLMs on the market are all very similar. Yes, there are differences and, yes, performance changes with complex tasks. But in the awards context, model choice is more about regional availability, personal preferences and how many tokens it consumes.
All to say, don’t centre on a model—centre on the task at hand. What job do you want Award Force AI tools to perform?
Consider an entry summary “job”. You want AI to capture the important entry information as accurately as possible, avoid embellishment and present it back to you in a predictable format.
That instantly gives you several things to configure. For example, if you want to summarise your awards entries
You might call that model configuration “Submission summary” and place it wherever you need that summary to appear.
Now consider another job. Let’s say we want to identify themes across a group of entries. You might use the exact same underlying AI model, but give it different instructions and a different temperature because the “job” requires more interpretation from the AI.
The underlying technology didn’t change but the job did… and you can expect the AI-powered outcome (or output) to be materially different.
This sounds like marketing gobbledygook but the reality is, Award Force isn’t interested in the AI-hype game, we’re interested in giving our very human clients the flexibility to make AI work the way they need it, today.
We could have built a summary button or even an AI-judge—but is that everything you will need? Unlikely. So we gave you the power to easily create what you need, without making you pay for something you’ll never use right now. We will undoubtedly make straightforward AI tasks even easier with dedicated summary buttons, AI-powered judging tools and awards agents in future.
But for now, as this technology moves at the speed of light, we are giving our clients flexibility first, for their maximum benefit.
That flexibility and respect for your unique workflow is powered by what we refer to as the pre-prompt. A pre-prompt gives the AI instructions before the user’s own prompt is added. It can define the AI’s role, what it should pay attention to, what it should avoid and how its answer should be structured.
Without a pre-prompt, each time you use our AI tools, you’ll need to define exactly what you want the AI to do. With a well-designed pre-prompt (and temperature!), those instructions can already be built in And in the case of our summary example, you can just ask for a summary— “Please summarise”. You don’t need to prompt everything else again, the pre-prompt already includes it.
Pre-prompts save everyone time but the real benefit is consistency. Important instructions no longer depend on memory or sharing prompts across your team. The model configuration carries all those instructions with it.
While pre-prompts tell the AI what you want it to do, temperature helps influence how predictable the AI output will be. In simplest terms, lower temperatures tend to produce more consistent and conservative responses, whereas higher temperatures are more creative and allow for greater variation. Importantly, there is no universally correct temperature—it depends on the job you want it to do.
For entry summaries, calculations, information extraction or other tasks where consistency really matters, a lower temperature is best. For generating feedback, exploring themes or other creative tasks, greater variation (or creativity) is more useful.
Note: AI with a high temperature is given more creative license in generating its response. A more creative or exploratory response may be longer or take a less direct path to an answer, potentially increasing the number of output tokens consumed and therefore the costs associated with running each task. It’s not a wild swing between cheap and suddenly expensive, but it’s worth noting all the same.
Once you have several configurations, naming your configuration becomes surprisingly important. A label such as Claude or Mistral tells the user which technology is underneath the configuration, but it doesn’t explain why they should choose it.
Names such as “Submission summary”, “Theme finder” or “Score normalisation” make the intended purpose super clear and also help improve how our users think about using AI.
It also gives our clients flexibility behind the scenes. If a new model arrives or an existing model becomes available in their region, they can simply swap the model and leave their configuration—and task at hand—untouched.
It’s possible to put a lot of information in a pre-prompt but keep in mind more instructions are not necessarily better. A good configuration should give the model enough info to understand:
Instructions that have little to do with the job can make the configuration harder to understand, test and maintain. If your preprompt starts accumulating instructions for several unrelated jobs, that is often a sign you need another configuration.
Test the whole configuration, not just the prompt.
AI configurations should be treated like other parts of your awards workflow—test widely before rolling them out. Use several examples rather than one ideal entry, include long entries, short entries, unusual wording, missing information and any edge case entries. Then check and compare the results. Does the summary consistently focus on the right information? Does the score normalisation behave reliably? Does the communication assistant always use the right tone? If not, adjust the configuration and test again.
While we’ve gone to great lengths to impress that the actual underlying model is less important than the job—model selection still matters. Different AI models vary in capability, speed and cost. A relatively straightforward extraction task may not need the same model as a complex analysis of lengthy entries. Likewise, there may be little value in generating a long answer when the person using the confirmation only needs 5 bullet points.
For frequently repeated jobs, small differences in response length or model cost can become significant at scale. So again, it’s not quite about which model is best—it’s about what combination of model and settings gives you the quality, consistency, speed and cost for the job you need doing.
AI can help people process information faster, identify useful signals and reduce repetitive work. That doesn’t mean every awards job should be handed over to AI. Decisions involving eligibility, judging and communications carry reputational, ethical and governance implications. AI can support those processes by organising information, summarising evidence, surfacing scores, synthesising or drawing attention to relevant material… but people should remain responsible for consequential decisions. It’s good for everyone.
We don’t want to inadvertently have AI-crafted entries judged by AI-powered solutions; that’s a risk many AI-native providers are running headlong into without consideration for the very real humanity of what it means to be recognised for excellence.
The biggest opportunity with configurable awards AI is not having access to more models, it’s being able to create a collection of clearly defined tools for your awards program. One configuration for summaries, one to extract entry information, another for moderation, another for feedback, etc.
That makes AI easier for teams to use, test and govern. It also means the people running your awards program do not all need to become expert prompt engineers because much of that thinking was already done when the configuration was created.
This is where AI starts saving you time, money and energy. Not one general assistant that tries to do too many things, but rather a collection of AI tools, each designed for the job at hand.
What awards job can Award Force AI tools do for you today? Test it out and let us know!
Articles
Feature focus
How-to guides
Press releases
Product updates