As you probably know, at Pivot Labs, we're big supporters of dual-use emerging technology. But there's an issue here we have to talk about.

To us, “dual-use” means technology which is designed for use in civilian applications along with national security, critical infrastructure. Dual-use means there applications in keeping society safe. An obvious application is cybersecurity, but consider applications such as OSINT, defending against disinformation, hate-speech detection and protecting elections and you'll see where we're coming from.

Pitching emerging technology to organisations who operate to keep society safe can be complex, as they typically operate in high-compliance environments in response to some very challenging threats, and this can manifest itself as a seaming barrier to adoption of new technology. This is an issue that startup founders can find it very hard to navigate.

It is all too easy to bash government for slow, monolithic approaches to technology but that is not what this blog post is about.

Issues

So what are the potential issues that a startup founder will face?

• procurement processes can be complex and expensive to interact with: Government and public organisations are required to demonstrate fiscal controls that ensure fairness and the same level of transparency to everyone participating. This is challenging environment in which to operate for everyone involved. Procurement usually means competitive open tenders. In this environment, it is difficult to imagine how a startup with a unique emerging technology solution would ever be able to respond to a bid.
• research and innovation grants appear to be a neat way to bypass the above problem by allowing startups to bid for capital to take a unique solution to a higher TRL. The innovation grant process can attract many applications and so can suffer the same problems with the procurement processes described previously.
• false demand-signals are created by innovation grants and government accelerators. They can be perceived to be an indicator that government would procure a product which results from these investment processes, but this is rarely the case.
• high-compliance requirements e.g. security certification requirements which can make it hard for a small business to qualify.
• low-tolerance security architectures severe architectural constraints can make it hard for a startup to create a product which can be used in such environments. For example, there are government organisations which operate segregated networks. Such restrictions present a problem for startups wanting to create SaaS products or accelerate their time-to-market by using other cloud/internet SaaS services.

Environment divergence

The latter point is mind-blowing to somebody who hasn’t appreciated the complexity of such an environment. Government departments and large public organisations have technology usage patterns that differentiate quite significantly from “the norm”.

Startup founders have enough of a problem - they are typically going to pick the best technology with the lowest barrier to entry to accelerate a product to market. As a founder you will approach this problem with a whole heap of assumptions about services which are ‘just there’ such as Google/AWS cloud services, public package repositories, AI/ML models on HuggingFace, source code repositories and software build services on Github.

Talking of AI, the landscape is going to look profoundly different on a closed network:

• The most effective commercial LLMs are supported by infrastructure and scale which do not make it possible to consider self-build alternatives. Is is far, far, beyond the scope of most organisations to deploy something that has the same efficacy as the latest ChatGPT, VertexAI or Anthropic offerings.
• SaaS and open-source models which you might consider for self-hosting are updated regularly and should not be considered static.
• Anything that you would consider for self-hosting is going to have demanding infrastructure requirements. Startups looking to self-host on high-end GPU hardware have joined a market where there is considerably more demand than supply. In a nutshell, it is impossible to imagine mainstream AI capabilities in use on a segregated network.

Strategies

So, what strategies can help here?

• Build for the future that doesn’t exist yet: The pace of change on the internet is forcing everyone to do business at internet speed without compromise. Build for that future.
• Ignore dual-use: As a startup founder you already have enough pressure to deal with. Ignore the problems you can’t solve and explore the marketplace which wants to deal with you.
• Keep the dream alive: So it may not be possible to market to government or public bodies in the early stages of your business development. Opportunities will arise further down the line, but you need to be realistic that it will be beyond series-A/B investment. Work out how to keep the dream alive and pitch the dream to investors.
• Find supply chains: Your early stage startup can't do business with your target dual-use partners, so work out the network which can take you there indirectly. There are many kinds of entity which can take your product and service and open it up to a wider market e.g. resellers, distributors, VAR. In the defence sector "Primes" have a role to foster collaboration and innovation across the sector. Engaging with a particular incubator or accelerator may also open supply chain options for your early-stage business now or further down the line.
• Architect for multiple environments: This is where the CTO or chief engineer earns their money. Don’t just build for AWS. Build for many environments: build using containers because containers can be deployed in different environments. Use automated deployment scripts which can be ported to different environments. Don’t fix all the storage or service naming around your first deployment target, and envisage components being deployment in many different ways. Consider building in more than one cloud environment as a simple test of how portable your architecture can be.
• Be an advocate for the change you want to see: As a dual-use startup, you are marketing your vision around a future which doesn’t exist, you need to clearly articulate that future looks like and bring that into your vision.