Why Should I use a grant writer?

Innovate UK grants are highly competitive, and putting together an application that stands a good chance of winning represents a significant time commitment. So if you’re going to apply, you’ll want to maximise your chances of success. There are some good reasons why you may want to consider contracting a grant-writer to help with this process. In addition to reducing the time investment required, an experienced grant-writer will understand what the assessors who will score the proposal expect to see in a bid, and will be able to provide guidance as to what is appropriate for the scope of a project. They also provide an external perspective, which can help ensure the bid represents a complete description of the innovation that is understandable to someone with no knowledge of the company or product.

Understanding the audience

I have 8 years’ experience writing grants for dozens of clients, mostly targeting Innovate UK. One of the benefits of Innovate UK’s assessment process is the detailed written feedback from the assessors. I’ve analysed the feedback from around a hundred proposals (so ~500 sets of assessor feedback, as there are typically 5 of them) across multiple grants competitions, giving me a good appreciation of what the assessors are looking for.

This provides good insight into the level of detail that they expect to see in things such as the characterisation of the target market and the project plan, as well as what costs are generally considered reasonable. It also gives good insight into the level of technical detail to include in a proposal. This is particularly important, because ultimately the proposals are for R&D projects and therefore must convey to the assessors that there is a substantial technical work programme.

Innovate UK assessors are a diverse group of people. Not only do they self-declare their technical expertise, but also Innovate UK has a finite pool of people to draw from. Therefore your audience is likely to include experts seeking in-depth technical information, as well as people unfamiliar with the field or any domain-specific jargon, who may struggle to follow details of the technology. Part of the art of putting together a winning proposal is pitching it with an appropriate level of technical detail to appeal to this varied audience.

Scope

There are two elements to scope issues: (1) whether the proposed project is a good fit for the competition and (2) how well the scale of the proposed project fits with what assessors typically seek in a proposal. The first is critical because no matter how good a proposal is, if an assessor concludes it’s a poor scope fit, then they are unlikely to recommend it for funding. Innovate UK publishes the competition scope and usually holds a briefing event (which they generally have a good grace to publish online so that you don’t have to attend the 2-hour-long video call). Experience is valuable here, however, as the language used can be open to interpretation, and it’s not always entirely obvious how a proposed project will align with what the assessors will expect to see based on the scope.

The second important consideration is the breadth of the project work program. Innovate UK funds R&D projects, not companies; while considerations such as the team and product-market fit are clearly important, above all the assessors will want to see a coherent program of R&D that addresses a specific problem or technical challenge. This is critical because arguably the most important aspect of writing a winning proposal is a clear description of the limitations of existing solutions to the problem you seek to address, and how the proposed innovation improves over them. The proposal therefore must focus on a well-defined, singular problem space, otherwise you will need to describe multiple problems and the various existing solutions, together with their limitations and how the proposed project will improve over them (just not feasible given the space constraints and the limited attention of the assessors).

An external perspective

It’s vital that the reader understands the core of the problem that you are setting out to solve, and appreciates the motivation for investing taxpayer money to implement the project. There are some important points that must come across early on in the proposal:

The problem that the project will solve for the users of the product or service it creates. The assessors may well be entirely unfamiliar with the area in which the business operates, and even if they are, they will appreciate the clarity.
The existing solutions to this problem, and why they are unsatisfactory. Assessors likely have little knowledge of the current state of the art, and so you will need to spell out to them why existing solutions are not good enough.
The challenging technical problem that the project addresses. You’re asking for government support, typically because what you’re proposing entails significant technological risk and hence investors may be reluctant to support the project in full. So the assessors need to be convinced that there is a substantial technical work program, which entails significant novelty.

These above three points need to be made in sufficient detail that the assessors will really get it, yet must be understandable to readers with limited technical and/or commercial knowledge of the area in which the company operates. It must strike the balance of fully describing a complex technical innovation while also being understandable to a non-technical audience. In my experience, the most common reasons that bids fail are because assessors have failed to understand an important aspect of the proposal. Working with a sufficiently thorough and inquisitive external party can add significant value, because if the project makes sense to the grant-writer, there’s every chance it will also make sense to the assessors.

Can't I just get ChatGPT to write if for me?

ChatGPT has an extremely broad knowledge base and is great at producing prose. Further, grant awarding bodies such as Innovate UK provides detailed guidance to applicants explaining what they want to see in a proposal. So one may expect that by feeding this into ChatGPT, combined with an outline of the proposal, surely it ought to be able to generate a well-written proposal that ticks all the right boxes?

Well, not exactly... In practice the results of getting ChatGPT to produce a complete draft of anything are underwhelming. First, it’s difficult to get ChatGPT to produce materials with the appropriate tone and formatting; rather it tends to generate prose in a particular style, and when you try and tell it how you would like it to change the style or formatting, it becomes almost recalcitrant in its failure to do so. In principle, this limitation can be addressed by fine-tuning the model based on a number of examples (which implies use of the API in the case of OpenAI’s GPT models, rather than the ChatGPT web interface); however, this rather undermines the value of getting it to produce content for you quickly and easily.

The most critical limitation of ChatGPT, however, is not in terms of tone and formatting; rather, it stems from the fundamental understanding—or lack thereof—of the content it generates. LLMs such as ChatGPT are based on statistical relationships between words, learned from a training distribution. They lack the higher-level conceptual frameworks that humans possess, making them deficient in areas such as complex reasoning and contextual comprehension (which partly explains why LLMs are prone to making logical mistakes and hallucinating false or spurious information).

Authoring an innovation proposal is a complex task that entails a comprehensive grasp of various technical, business, and strategic aspects. Further, the assessment process is not just a box-ticking exercise. Putting together a compelling proposal requires crafting a persuasive argument that substantiates a technological innovation's ability to address real-world problems. This means that the author needs a real understanding of multiple pertinent aspects, enabling them to present a cogent argument for a technological innovation that solves a real problem and evidences understanding of user-need, resulting in a convincing business case that ties in with a coherent technical approach and work program.

Assessors interpret the guidance based on years of experience, whereas an LLM lacks this kind of contextual understanding. To put things another way, it’s entirely possible to create a proposal that interprets the guidance literally and ticks all the boxes, yet an experienced human will conclude that it’s fundamentally just a bad idea and a waste of taxpayer money. LLMs may excel at generating well-written text, but they lack the cognitive capacity and contextual awareness that is required to create a coherent proposal that will resonate with assessors, and so extensive human oversight is required to ensure the LLM is generating appropriate materials.

I have come to see ChatGPT as a productivity tool, rather than a way to automate complex tasks. For example, it’s great at summarising ASR-generated call transcripts. It’s also good at proofreading and useful for providing suggestions to reword things to avoid repetition or reduce word count, etc. It’s even useful for providing constructive criticisms—albeit with results that are somewhat hit-and-miss. It’s a powerful tool that’s very good at automating relatively straightforward tasks; however, it lacks the capability to be able to direct itself to solve complex problems and so it still needs a human in the driving seat to tell it what to do and check that the results are meaningful.