At Elula our core business is AI/machine learning. We are solely focused on building AI software services to improve customer engagement. This is key to our success. However, many big corporates struggle to extract value from their AI/ML function. Common complaints include:
- projects being unable to move beyond pilot stage;
- projects being deprioritised because data is not core to the business;
- data scientists misunderstanding the business;
- data scientists being disconnected from organisational priorities and strategy; and
- data science teams functioning as research teams—instead of commercial.
Cognisant of these pitfalls, Elula focuses our data scientists’ work on solving big, important problems for our customers that most importantly deliver outcomes. We do this by:
- propagating a unified, company-wide understanding of business goals;
- throttling ad hoc analyses (so common among data teams);
- strictly advocating the build of scalable, repeatable AI products;
- giving different teams shared business goals;
- prioritising continual product development toward customer needs; and
- maintaining an appropriate, competitive tech stack.
Unified business awareness and outcomes
Everyone at Elula develops the same suite of interconnected AI software products, which yields major benefits. First, everyone has overlapping experience with the same products, simplifying learning curves for all. Naturally, the same people are also involved throughout development and testing, so they find and fix gaps first-hand instead of writing work tickets for distant teams.
Second, all employees are aligned to common goals and visions—all striving to achieve equal outcomes. Everyone knows what each team is developing, enabling company-wide awareness of our big-picture strategy and goals. This fosters a general feeling where people know they’re part of things bigger than their individual projects.
Third, all teams have similar priorities, thus incentivising collaboration. In many big corporates, departments need each other but have separate objectives, which disappointingly deprioritises collaboration.
Elula’s organisational structure enhances these three points. Rather than having separate teams for separate functions (e.g., the engineers being organisationally separate from the data scientists), Elula aligns its teams by objectives: the product engineers and product data scientists, as opposed to their client delivery counterparts, all report to our product head, ensuring common, unified objectives.
Minimal ad hoc
Most data professionals deal with ad hoc analysis requests throughout their work on major data initiatives. For example, data scientists developing price optimisers might be asked to gather numbers for the sales department; the projects are unrelated, but organisations often come to rely on data scientists for anything data-related—without considering long-term business consequences. Even more common is data scientists producing monthly reports for any departments that ask; it’s important work, but it doesn’t require an AI team and it shouldn’t distract from AI business goals. Really, companies without established metrics reporting are rarely well-positioned to extract business value from AI teams.
At Elula, ad hoc analyses still happen—our customers come first—but they’re fielded by a dedicated customer delivery team whose primary responsibilities include such work; this alleviates the ad hoc burden on others. We triage less-urgent ad hoc requests through our product development team so that major initiatives suffer minimal disruption.
In addition to ad hoc requests, most data professionals are burdened with the familiar labour-intensive “business-as-usual” tasks: important, but repetitive parts of major products/processes. For example, data engineers might handle regular batch updates for three hours every week. Add enough of these tasks, and soon you have highly qualified, and expensive, staff doing repetitive, automatable maintenance work.
Elula proudly automates major processes and tasks. In fact, we explicitly avoid traditional AI consulting business because it’s famously unscalable. Scalability is a company-wide emphasis.
In addition to being good business, this scalability enables us to provide much more to our customers: 1) we deliver products more quickly, 2) we have fewer human errors, and 3) we apply more of our focus and expertise to customers’ primary business problems instead of implementation or logistical problems.
All critical professions together
Prioritising scalability requires experienced software engineers; but data scientists are rarely trained in writing stable, scalable, production-grade code. Nonetheless, in many companies, data and engineering teams are separated (for example, the data team might handle BI while engineers develop apps). This separation either leads to below-production-grade AI codebases (if the engineers take no part) or slow productionisation (if the engineers must “translate” code from the data team into production-grade).
As mentioned earlier, Elula’s data scientists, data engineers, and software engineers all work together toward the same business outcomes; the three teams regularly collaborate throughout the day. Our engineers build internal products that our data team uses for AI, enabling both data-scientist-grade AI and engineer-grade code.
Product research and development primacy
Most companies like the idea of building new products and features but AI research and development offers uncertain payoffs, making businesses wary or frustrated. It’s hard to get leadership approval for a claim like, “if we develop new adversarial neural networks, we might improve predictions by 10%” when decision-makers are looking to flatten or reduce budgets.
Product development has been one of Elula’s primary strategic goals since genesis; we’re always building something new. Although we face the same uncertainty everyone has with AI development, we believe that staying state-of-the-art is critical in this era of evolving AI technology. We want to know, not assume, that our products and services are top-of-the-line.
In addition to the obvious payoff of superior products, our customers also benefit from this philosophy which effectively spreads the cost of R&D.
Rapidly evolving tech stack
For our AI products to remain top-of-the-line, our tech stack needs to stay competitive and relevant. Because our engineering and data teams work so closely, Elula’s engineers have insight into what tools best serve our AI needs. And because product development is paramount, leadership is quick to approve new tools that demonstrably suit our needs. For example, when our data team needed an IDE that quickly identified metadata, while at the same time our software team wanted better testing capabilities, we quickly upgraded from PyCharm Community to Professional.
Companies often struggle with this process as organisations are wary of incurring the financial, labour, and training costs associated with new tooling—just getting people to use new tools can be arduous. For example, companies with deep bureaucracies and legacy systems will balk at cloud migration, or will require near-unanimous consensus on decisions, thus delaying time-to-adoption for new tech.
There are benefits to working for big corporates (e.g., stability, predictability, deep specialisation) but there are obvious drawbacks. What makes AI development at Elula special is that we deliberately avoid these drawbacks. We all share the same goals for the same product suite, we avoid ad hoc/non-core projects, scalability reigns, our critical AI professionals sit together in-house, product development never takes a back seat, and our tools keep pace with our goals.