Once you have a solid machine learning (ML) idea or question, how do you decide who is going to build and deploy the model?
There are three paths an organization can choose from when deciding who should develop their ML solution: building one in-house, buying an off-the-shelf solution from an existing vendor, or partnering with a vendor to build one.
Of course, there are pros and cons to each option. Let’s get to know them:
Building means an organization builds the ML solution in-house – from ideation and data processing right through to execution and deployment.
Building is often the most intensive option, where the onus is on the organization to plan, execute and evaluate every step. Those who choose to build in-house will need to invest in the upskilling of existing talent or hiring and onboarding of new talent, which makes this method more expensive up-front. As with any new project, an organization would need to develop new processes from scratch to manage ML projects.
However, it is often the most rewarding – the systems are fit to the organization’s specific systems and needs, and the solution can be tuned or adapted with changing requirements and conditions. The organization can also use the learnings to apply to future projects (and ramp up the difficulty/complexity over time), and its data and ML solutions will remain proprietary. And though it’s the most expensive option up-front, it is more cost-effective in the long term, particularly if an organization plans to integrate ML into its business practices and deploy multiple ML projects.
Another route is to buy an existing solution, usually in the form of an outright purchase of the solution or purchase by subscription.
Of the three options, buying is the lightest and least-expensive lift, designed to plug into your systems and provide immediate returns. Depending on the arrangement, continuous support, updates and improvements may be available for the product.
However, it lacks the fit and customizability of the other two options; the solution might not exactly address the problem that the organization wants to address. And though the returns are immediate, they would not be as profitable as the other two options.
An organization may also choose to partner with another company to create a customized ML solution. Typically, the organization will go to the partner company with an idea or problem, and the solution is then developed and deployed by this partner company.
Partnering is less intensive than building but more so than buying. It can result in an ML solution that is customized to the organization’s needs, with a lower up-front investment, as an organization would not need to upskill or hire additional talent. Organizations would also benefit from the expertise from specialists, often from multiple domains (business and technical) depending on the partner.
However, if a company wants ML to form a core part of its competitive advantage -- that is, if it's planning to work ML into a company's processes to continually create efficiencies across the organization -- partnering is not an ideal option. The cost of partnering on individual projects will eventually be more expensive than building in-house, and it would be extremely challenging to maintain the level and quality of communication required.
Each ML problem is unique; different problems and strategic needs will fit better with different pathways. We encourage organizations to deeply consider how ML fits in their business model and how important having the knowledge and capability in-house is to the organization.
Amii works with organizations that want to build their own ML solutions in-house. We help organizations train, upskill and hire technical teams; teach non-technical employees foundational ML definitions and concepts; and guide organizations through the entire AI adoption process, from teaching them how to define an ML question to the development and execution of their ML solution.
This method de-risks the building process. With Amii, teams have a built-in guidance system for these complicated projects and aren’t left spinning their wheels when they encounter issues and roadblocks.
It also leaves the organization in good shape to pursue ML after the engagement: giving them the tools, vocabulary and experience to experiment and take on new projects. Ideally, an organization will get into a state where they are able to constantly increase efficiencies – some even open up new revenue streams by selling proprietary solutions to other organizations.
When deciding whether to build, buy or partner, we recommend having a clear understanding of your organization's expectations and strategic goals -- you don't need to go in knowing how to use AI, but it's important to understand what you want AI to do for your company.
Nov 22nd 2023
On October 17, Dr. Deepa Krishnaswamy, from Boston's Brigham and Women's Hospital, presented “AI-derived annotations for lung cancer collections using NCI Imaging Data Commons" at the AI Seminar.
Nov 16th 2023
Thinking of pursuing a PhD? Amii researcher Alona Fyshe explores how undertaking a PhD is like training for the Olympics, and how you can develop an Olympian state of mind.
Nov 16th 2023
Read our November monthly update on Alberta’s AI ecosystem. Featured this month: Impact Report & New U of A Faculty Positions.
Looking to build AI capacity? Need a speaker at your event?
Get involved in Alberta's growing AI ecosystem! Speaker, sponsorship, and letter of support requests welcome.
Curious about study options under one of our researchers? Want more information on training opportunities?