Building your data science team is one of the most essential elements of developing your AI tool. Naturally, an AI project doesn’t start with development right away, but with the right preparation – like creating a very basic data strategy including the use case, information about data, etc., there comes a point where you have to get yourself a data science team. And what then? Do you start recruiting? Outsource? What’s best?
Let’s have a look at some pros and cons of both solutions.
In-house data science
For many companies, choosing to build an in-house data science team seems like the only option out there – and that’s mainly because the bigger players like Google, Facebook, Netflix have their own teams. When you’re a big company, it’s definitely a good idea to start building a new team of data scientists, however, for many small and medium companies, it’s often not an available solution.
So why is an in-house team good?
Pros of in-house data science
When you have your own team, you decide where to go. If you need to make some changes, they’re done fast, if you need to talk to your data science, you have them right there. If you need some level of customization of your solution, an in-house data science team may handle that better. They’re a part of your company, so the assumption is that they understand your business and thus can create a better-customized solution.
However, it’s important to understand that customization is not such a huge challenge to cross outsourcing off the list. Additionally, if you’re hiring your own team and starting from scratch, the understanding of your overall business is also not there…
Your AI solution is also an asset of your company. When building an AI solution with an in-house team, it’s clear that everything that’s created is yours. Intellectual property may be especially important if your solution is something you offer to your customers – say as a SaaS product, or when it’s something new and promising. With an in-house data science team, managing intellectual property is easier, but in an outsourced team, you can also maintain the property of the developed solution as long as that’s stated in your contract.
If you simply don’t have to rely on anyone, you’re independent and free. You can work on as many features as you wish, you communicate straight with your team, and you don’t depend on somebody’s knowledge and experience. Isn’t that wonderful? There will be no time zone difference (unless you work with remote teams) and no language barrier (unless you build an international team). An in-house team working in your office is a straightforward solution that gives you a lot of independence and comfort.
Cons of in-house data science
Time and money
It is estimated that hiring a data scientist can take over 20 months and cost 15 thousand dollars in recruitment and hiring costs. Yes, that’s on top of their salary – which is a decent amount of money, too. Mind that hiring one data scientist doesn’t solve the problem – you can’t rely on a single person to do the work of a team – and then, when such a person quits, you’re left with nothing. And the process starts all over again. Hiring data scientists is a time-consuming and expensive process, so it only makes sense when you’re fully committed to maintaining the team long-term.
Difficulty assessing skills
If you’re just starting with AI and you’re not a data scientist yourself, how do you know the person you’re interviewing is an actual AI expert? Before you get to the interview, how do you even know what skills you’re looking for to develop your project? Without the technical know-how, you won’t be able to accurately assess the skills of your candidates. So it’s either up to luck, or you can hire a consultant – but then you’re turning to external specialists anyway.
Lack of focus on delivery
When you’re working with an external partner, you’ve got a contract that states that they must deliver X, Y, and Z. Usually within a given timeframe. Naturally, there’s room for adjustments and changes, but generally – everything is well established at the very beginning. It’s also to their best interest to deliver your project as soon as they can to have a happy customer and move on to other tasks. With an in-house team, however, there isn’t as much pressure on delivery. I don’t want to say that an in-house team’s work is slower, that clearly depends on various factors, but in general, an in-house team doesn’t emphasize the importance of acting fast as much. This case, however, is related to teamwork and project management and can be solved if approached properly.
Scarcity of experts
The demand for AI-related roles is much higher than the supply. In a Racounter article, Dr. Paula Parpart, the founder of Brainpool, said:
Demand for top AI experts and data scientists is far outstripping supply, which is why outsourcing is a compelling option. Some of the specialisms required are so niche, the talent so hard to find, and contractual relationships so tight that it could take two years for an organization to fill a role.
So it’s one thing that there aren’t enough AI specialists in general, and another thing that there are even fewer experts who will have both the right technical skills and understanding of your business to apply the right tech to your particular business problem.
What’s more, companies are fighting to attract top talent and organizations outside the very exclusive FAMGA group (Facebook, Amazon, Microsoft, Google, Apple) don’t have instant and unlimited access to talented data scientists. It’s difficult to find the right people – and then, you might not even know who the “right” people are.
Outsourcing data science
More and more organizations turn to outsourcing to start their AI journey. It’s become a solution recommended for the first AI projects, before organizations are ready to go company-wide with AI adoption. In Andrew Ng’s AI Transformation Playbook, you’ll read:
While outsourced partners with deep technical AI expertise can help you gain that initial momentum faster, in the long term it will be more efficient to execute some projects with an in-house AI team.
Let’s have a closer look at some pros and cons of outsourcing data science.
Pros of outsourcing data science
Starting right away
That’s one of the biggest advantages of outsourcing data science. When you get an external partner to work on your AI project, you don’t need years of recruitment. Sure, you need some research to shortlist the companies you consider, you need some time for the preliminary talks to make sure it’s a good match – but overall, you save months. Starting faster is often crucial because you save money that you would be spending on ongoing recruitment processes, and you can stay a step ahead of your competition. After all, time is money, so waiting months to start working on your AI project is just way too long.
When looking for an external partner, you can verify their experience. Check their portfolio, read case studies, have a look at their team members. It’s a good idea to go through the content they create to see whether they meet your expectations. Can they deliver what you need? Do you feel you might work effectively together? It’s both about the good match between two businesses, but also about that spark – when you feel you will get on well.
AI for business is still in its infancy and some companies may take advantage of this situation trying to lure you with “best services” at “best prices”. You have to dig deeper and look for proof that they can deliver – and if not with case studies and content, go to trusted sources like Clutch where software development companies are reviewed by their clients and ranked in given categories.
Seems like something obvious, yet it’s a huge advantage. A B2B contract between you and your development partner is different than a standard contract of employment. It allows you to establish a timeframe and budget for the project – estimated by the contractor and acceptable for both parties. What’s in the contract depends on your contractor and your individual needs, but it’s worth remembering that it’s a legal document protecting you in case something goes wrong. If something goes wrong in your in-house team, you can fire your employee and start over again.
Opportunity to work with experts
Outsourcing companies usually hire people from various backgrounds who have experience with different technologies and projects. They offer the support of experts that can work for you or just advise you when necessary. And even though their hourly rates are not low (sorry, I can’t help you find inexpensive AI development), the value lies in the efficiency. They’re experienced, so they know how to help and don’t beat around the bush. They use the time they have for you very productively. And let’s be honest, you wouldn’t be able to hire all these experts from different fields in-house. In comparison to having an in-house data science team, outsourcing can be more cost-efficient.
Since outsourcing companies gather people with different professional experiences, they will be able to advise you on the right solution, good preparation, and so on. Even if you have zero experience with AI projects, a good external partner should be able to guide you through the process of AI adoption. It goes far beyond model building: it starts with matching the business and tech, building the use case, creating a data strategy, preparing the data. What’s more, given that an external company has wide experience with other projects, probably also similar to yours, they will be able to show you what challenges you might face along the way and how to deal with them, as well as point common mistakes other organizations make – to help you save time and money.
Assuming that you hire a company that’s experienced, you will benefit from their expertise in a number of ways: they will guide you through the process, teach you how to work with AI projects (even if all you do is watch what happens – but you do want to get involved!), and deliver results faster. And they’re faster for 2 reasons: 1) they know how to do it, and 2) they’re delivery-oriented. You’re their client – they’re committed to delivering value as fast as possible to make sure that there are tangible results for your business. They don’t work just to develop a cool model. They work to help you save or make more money.
Since you don’t wait long to hire a team, and you reach the goal faster, you save money. Additionally, an outsourcing agency allows you to change the size of your team depending on your needs. Is the project going great? Cool, add more engineers (or other roles) and make the most out of it. Are you unsure where you’re going or need to slow down? Make the team smaller or pause your project. You wouldn’t be able to do that with an in-house data science team – if you let them go, it will be hard to win them back after some time.
Cons of outsourcing data
Lack of domain expertise
When looking for an outsourcing company, you’re not sure they’ve got the domain expertise you need. Whatever your core business is, it is essential to your AI solution. AI is just another tool to help you achieve your objectives, not a goal on its own. So, you need your team to understand your business. It may happen that you get a team of great AI experts but they don’t understand your company, your pains, and the processes you’re trying to improve. To avoid such a problem, and it’s a serious problem, you need to communicate. Make sure everything is clear to everyone involved in the work on your AI solution (both on your side and your contractor’s), ask and answer questions, map processes, write things down. You can’t really expect that your outsourced team will be a collection of omniscient people – but you can teach them what matters.
At Neoteric, we solved this issue by starting AI projects with a 2-day AI Sprint workshop. During the workshop, we cover a variety of topics related to the business, the tech, and the project itself. Both domain experts and data scientists take part in the workshop to allow for a deep understanding on both sides and smooth communication between teams.
Giving away control
Working with a team that’s not “yours” may feel like you’re giving away control. You’re not fully in charge of the project, you don’t watch over every step of the process. But that’s a good thing! You don’t have to be involved in every action there is, make a list of action items and take care of those assigned to you. Schedule calls to catch up on the progress and keep in touch with your team. That’s all the control you need. It may feel as if you’re handing a piece of your business over to another company, and it might be uncomfortable at first, but it’s something you need to learn and adjust to.
Difficulty finding a good development partner
Just like in the case of hiring an in-house team, when you have no experience with AI, you may have difficulties finding the right team. There’s a whole bunch of companies claiming to do AI – but what does this AI do? Is that what you need? The thing is that your contractor shouldn’t expect you to come to them with a ready data strategy and everything pre-established. You need them for guidance, consulting, brainstorming – and then for the AI work. However, it may be difficult to evaluate how right (or wrong) a company is for you if you don’t know them. Learn more about them, meet them or call them – make sure you’ve got a good level of understanding of what your organization needs.
In-house vs. outsourced data science – what’s the winner?
The answer is always the same: it depends. You need to think about your needs, expectations, and possibilities. Do you have a lot of time? Do you want to go long-term with your AI team? Do you need quick wins? Depending on what you identify as the factors most important to your business, you may choose in-house or outsourcing as the best solution for you. If you’re looking for a team that will introduce you to the crazy world of AI wonders, help you start fast, and deliver results quickly – outsourcing is the way to go.
Learn how to use predictive analytics to drive business results and revenue