The journey of Ox Intel; from H4MoD to start-up

This blog was originally published by the Common Mission Project. We wanted to share it as it charts the progress of Ox Intel, a company that came out of the the first Hacking for MoD cohort at King's College London, which we covered in a post exactly a year ago today.

In an evolving security landscape, government agencies need to be creative and resourceful in developing solutions to complex problems, which requires looking beyond the defence space. By working with universities, governments can tap into innovative approaches offered by students, who can provide outside-the-box solutions to existing problems.

Hacking for MoD (H4MoD) is an accredited MA course at the Department of War Studies at King’s College London that pairs teams of postgraduate students with sponsors from the UK Ministry of Defence to tackle critical national security challenges. Our team was one of four in the inaugural class in the Department.

Over the course of H4MoD, we learned to apply Lean Startup methods to a real-world national security problem with our UK government sponsor, RAF Leeming. Over the 10-week program, we conducted extensive outreach to understand and validate our problem, speaking to over 80 prospective customers and stakeholders, and designed a minimum viable product (MVP). Our H4MoD programme ended in March, but we are continuing to work with our sponsor and other members of the UK and US intelligence communities to further develop and deploy our solution and are continuing to apply the lean startup methods learned in H4MoD.

The first step in our journey was breaking down our sponsor’s problem statement:

“In an era where intelligence officers face the daunting task of analysing terabytes of data to inform national security decision-making, there is a need to systematically account for subjectivity in intelligence analysis.”

With this problem statement our team began looking at human intelligence (HUMINT) sources, largely evaluated by human analysts, who we found appeared to be constrained by the amount of data they are able to process. Concurrently, our sponsor had interest in applying Artificial Intelligence (AI) and Machine Learning (ML) to the intelligence cycle. For this reason, our team explored opportunities to automate components of the process to account for bias and relieve some of the burden on human analysts.

 To develop this knowledge, our team had to learn about the problem ecosystem and the people within it, identifying the key stakeholders and articulating how they are impacted by our problem set — a process known as beneficiary discovery.

In our initial beneficiary discovery with practitioners and external technical experts, the team began testing the waters around our automation-focused approach. Despite strong appetite for the application of AI and ML among our beneficiaries, this discovery process revealed several barriers to automation — primarily the necessity of the human analyst’s judgment and the consensus that intelligence is an art, not a science that can be taught to machines. With a third of the term gone and team morale quite low, we knew it was time to change direction.

In the spirit of H4MoD, the best way of gaining more knowledge around the problem was getting out of the building and talking to people. We ramped up our outreach, averaging well over 10 interviews a week in the second half of the course in order to learn more about the problem ecosystem. These conversations identified our key beneficiaries’ pain points with the existing grading system for evaluating source and information reliability, which included a lack of consistency, transparency, and granularity. This helped to validate the decision by our team to pivot. While we were still interested in incorporating automation, the primary focus became designing a solution that provides a more uniform framework for intelligence analysis based on established indicators of reliability.

This line of inquiry would inform the development of our Ox MVP, which collects data on individual human judgments to generate insights based on collective wisdom. In addition to our beneficiaries, we also consulted data scientists, informatics specialists, applied mathematicians, and instructors of intelligence analysis. To address our beneficiary pain points, we built our MVP to enhance analytical rigour, support dynamic data collection, and power decision advantage across the intelligence cycle. We then set out to validate our MVP with the key beneficiaries who expressed real excitement in the potential of our solution, both in solving existing problems and unlocking opportunities around better understanding human judgments. This validation meant we had successfully reached product-mission fit, a milestone in the Lean Startup journey wherein our value proposition matches the needs of our primary beneficiaries.

We concluded H4MoD with a final presentation of our MVP, articulating our mission achievement to an audience of beneficiaries across the MoD. Since this presentation, we have generated sustained interest in our MVP, and are continuing to work with the beneficiaries identified through H4MoD as we move beyond the academic environment. We have since transitioned into the world of business, forming our startup, Ox Intel. 

At the outset of H4MoD, our team hoped to blend KCL’s high-caliber academic offerings with some practical problem-solving experience, gaining a skill set not typically taught in conventional MA courses. In the end, not only did the course successfully provide this unique learning opportunity, it also offered our team the chance to work directly with high-level practitioners in our chosen field and transformed into an opportunity to create meaningful change. Although none of us had worked together previously, our group dynamic quickly became one of our strongest assets. As the weeks went by, our strength as a unit grew, sustained by our pace of progress, as well as the feedback and validation from our beneficiaries. 

The journey is far from over and has not been without its challenges. These ranged from difficulties reaching end-users, to designing an effective deployment strategy, to being forced to move all operations online due to COVID-19. The adaptability we have learned in the process is something we have taken beyond H4MoD as we navigate the obstacles that stand in the way of successful deployment.  Moving forward, the team continues to lean on the support of our sponsor and the network of beneficiaries we built throughout the course. As Ox Intel embarks on the next phase of our journey, we will also continue to draw on the Lean Startup methodology to overcome challenges, and to build technology that can transform the art of intelligence.