Running and growing an AI startup comes with its own unique set of challenges compared to other sectors, ranging from; talent acquisition, how to implement the technology, busting the AI hype around what it is and its capabilities, the accessibility of trusted or secure data and a lack of AI expertise in industries.
In the first article of a three part series focusing on what founders are doing to navigate the fast growth AI industry, Information Age spoke to the founders of some the UK’s leading AI startups to understand the ten main pain points of running an AI business.
1. Talent acquisition
A major challenge for any growing AI startup, as well other technology-focused roles (like cyber security), is talent acquisition. In these industries there is often a talent shortfall compared to the number of jobs available.
Miriam Cha, co-founder and COO of Rahko — a quantum machine learning company focused on the discovery of new drugs and materials, explains that attracting the best people from relatively scarce talent pools is a major hurdle.
“As an AI startup you compete with much larger companies who have much deeper pockets than you so you need to offer something more important than salary. At Rahko, we have a friendly, fun team, but far and away the most attractive thing about joining us is how absolutely cutting edge our work is — if our work was boring we would have a very hard time convincing anyone to join us,” she says.
Darko Matovski, CEO and co-founder of causaLens — an AI platform that helps businesses improve KPIs, agrees that as an emerging discipline, “attracting top people with a diverse skill set to ensure we not only develop our scientific capabilities and products, but also provide the best service to our customers” is a major paint point.
“There is a high demand for AI expertise, so hiring in this sector is very competitive,” he adds.
However, Matovski suggests AI startups should view this as an opportunity. The best talent will only want to work in an environment with an “altruistic working culture and shared goals”.
Chris Ganje, CEO and founder at AMPLYFI — which uses AI analytics to mitigate emerging threats as they happen, suggests the challenge lies in the range of skills required at an AI startup.
He says: “It is not enough to lock some AI experts in a room. Engineering, data mining, user experience, design, digital marketing… a range of skills are required in the mix to really achieve a meaningful solution and value.”
Safe Hammad, CTO and co-founder at Arctic Shores — which improves the hiring process through behaviour-based assessment, agrees and says he has struggled to hire software engineers because they’re in such high demand.
However, he goes one step further. “It’s not enough to find talented people — we’re looking for those who truly align with our values too,” he adds.
2. Implementing the technology with existing infrastructure
The introduction of AI into a business’ strategy or product is still at a relatively early stage. And, while the potential benefits are clear — speeding up process or leveraging smart data for customer success — the challenge is “pinpointing precisely how, where and why you’re going to implement the technology for optimum results,” according to Dr Alex Young, founder and CEO at Virti — the enterprise AI learning solution.
He advises businesses must be clear on what outcomes they want to achieve, before incorporating AI into a business strategy or product.
Ky Nichol, CEO at Cutover — the work orchestration and observability platform, explains that this AI implementation pain point comes from some parts of a business that don’t have a “healthy consideration of human/machine collaboration, and so adoption of automation can often be tricky. It remains fairly well understood by technical teams, but as it moves across organisational requirements, there can be knowledge gaps between adoption and implementation.”
He continues: “As organisations design strategies to unlock this potential and complement automation with existing technology infrastructure, there will always be challenges to leverage existing investments too. The challenge now is to join up tools and build processes and flows of work across them.”
Ganje from AMPLYFI suggests that some companies and their infrastructure is simply not ready for AI — presenting another challenge for AI startups.
He says there is a need for “mature digital leadership to really exploit the benefits of AI in an organisation…[but] some leaders and companies are just not ready to use machines to enhance their decision making.”
3. A lack of AI expertise
As an extension to the above, one of the main pain points for AI startups in delivering their solution to market is a lack of AI expertise in the many industries that are trying to implement AI.
“Because AI has clear limitations that different industries are not familiar with, they view it as this all mighty capability that you can just turn on. The problems come in as they are not familiar with the foundations needed to run AI properly in their business. They are also not set up correctly to adapt to these requirements,” explains Ofri Ben-Porat, CEO at Edgify — a company that trains deep learning models at the edge.
“Stay focused on the business outcome, not doing bleeding edge stuff that does not help your customers” — Peter Ryding, founder, VIC (the AI coaching tool)
4. The data challenge
Matthew Hodgson, CEO and founder of Mosaic Smart Data — a startup that delivers the insight and real-time intelligence for the financial industry, highlights the irony that data resembles both the biggest challenge and most integral component to deploying AI effectively for any startup or business.
Looking specifically at financial institutions, Hodgson says that they must ensure that their data foundations are fit for purpose.
“Data is the raw material of our industry, and without it, the benefits and potential of AI are stunted and capped before the system even gets switched on. Many financial institutions already sit atop mountains of their own data in addition to buying more from vendors — yet they do not have the time, the resources or the staff expertise to sift through it,” Hodgson explains.
Dr Richard Ahlfeld, founder and CEO at Monolith AI — a startup that builds new machine learning software to help engineers to improve the product development process, echoes this view.
He says: “Any pain points tend to boil down to the data: getting the data, ensuring data security, making sure that you can trust the data.
“There’s no standardisation of what makes data ‘valuable’ across the industry either, and not all engineers follow the same protocols and practices. For example, deciding what data to keep can be tricky as it’s hard to anticipate what might or might not be useful to have in the future. Even saving data from failed ventures (a practice which is often overlooked) can have its value, as it acts as a reference for future experiments.”
A guide to artificial intelligence in enterprise: Is it right for your business?
5. Building trust
“One of the main pain points in running an AI business is counteracting misunderstandings people may have around AI and its capabilities,” says Mark Nicholson, CEO, Vivacity Labs — which improves traffic insights with artificial intelligence.
There is an inherent distrust in the technology and “overcoming the negative perceptions that some people have when it comes to AI, especially among a more traditional demographic,” is a significant challenge according to Philip Marshman, founder and CEO at Sentai — a startup using advanced AI to support informal caregivers look after the elderly.
To overcome this, Nicholson and Marshman have put data privacy and trust at the heart of their company’s product development — “and that should be the first thing entrepreneurs consider when setting up a tech/AI company,” according to Nicholson.
6. AI hype
In 2019, London-based investment firm MMC Ventures, in association with Barclays, analysed 2,830 AI start-ups in Europe and based their findings on public information and interviews with executives. It found that 40% of these companies did not actually use AI — a huge problem for startups is the amount of AI hype in the market.
Cha from Rahko explains that tempering expectations through hype is always going to be tricky.
“While fundamental advances in AI are always exciting and important, a lot of the critical work in graduating from hype is in making things work in the real world, which comes with its own set of challenges. This is doubly true for quantum computing. Quantum computing captures people’s imaginations which makes for a field that is very prone to wild hype. We are extremely excited about the powerful new capabilities that quantum computing will bring, but like with AI, we do not focus on technological advances that do not strongly seem like they will be able to solve a real, currently intractable problem.”
Hammad from Arctic Shores agrees and says that “it’s all too easy to create a product that’s great in theory but doesn’t have its users at heart. One challenge is to constantly keep the customer’s challenges in mind — which problems are you solving?”
7. Increased competition
AI is a fast growing industry and competition is fierce.
Daniel Cooper, managing director at Lolly Co — the digital transformation partner, sees the increasing amounts of competition on the web a significant pain point for AI startups.
“SEO has become hyper-competitive, even when factoring in individual niches. Getting noticed on the internet is becoming increasingly difficult, and is driving a return to more traditional PR methods to grab Google’s attention. Simply put, there is always someone who is getting higher traffic than you,” he says.
8. Covid-19
Every industry, even AI has been disrupted by Covid-19.
“From a technology perspective, AI businesses have had to learn and adapt quickly to ensure their technology can be implemented with minimal disruption, and while the pandemic was unexpected, adapting to these challenges has allowed many tech businesses to thrive,” says Tim Weil, CEO, Navenio — the indoor location solution.
Tom Reiss, CEO at Roby AI — the autonomous machine that empowers property managers, also says the pandemic has created a number of challenges for his company because of the impact on the sector that it serves — real estate. Like many businesses, “we’ve had to adapt quickly to our customers’ needs and the changing pace of the market,” he says.
9. Focus
Both Ben-Porat from Edgify and Hammad from Arctic Shores agree that focus is a challenge for any AI startup.
Ben-Porat explains: “Since AI seems like ‘magic’ to most low-tech industries (for example, retail, manufacturing, automotive, etcetera) it is difficult for them to focus on one key pain point that they are trying to solve with AI. In most cases, they also just want to say that they are using AI across the board, rather than focusing on one use case where they will be using AI to solve a specific problem.”
Hammad agrees but from a different angle. He argues “it’s far too easy to get sidetracked from building and improving your core product, either by darting from one product avenue to the next, or by trying to fit your product to a particular client. Do one thing, and do it well.”
10. The risk of radical change
AI has the potential to produce radical change in any organisation, if used correctly.
Pointing to the HR industry, Dr Alan Bourne, CEO and founder at Sova — the AI recruitment service, explains that “businesses can make huge long-term gains by automating elements of the recruitment process, like CV scanning and interview bookings. The biggest impact AI can bring to the HR industry is overhauling the fairness and accuracy of how people are selected for a job, by eliminating elements of inherent bias that hamper traditional, human led recruitment processes.”
However, he suggests that making radical changes using AI carries an inevitable risk versus payoff quandary.
“The stakes of challenging the status quo are huge, making the expectations much higher for the new solution to significantly outperform the original offering. Guiding people through the change process here needs to be based on genuine and thorough rigour, in order to offset decision makers’ risk levels.”
