In-house team vs machine learning outsourcing: what to choose?

Irina Kolesnikova
May 16th, 2022

A business leader choosing between in-house and outsourced machine learning teams

When starting a new project in the artificial intelligence field, apart from other important things to consider, business leaders often meet a dilemma: what is the best option – to hire an in-house team or choose machine learning outsourcing?

Considering the fact that 85% of AI projects fail, a wrong choice can cause significant financial losses. Therefore, we’ve interviewed startup founders who have already hit on the right solution. Check out their thoughts on the topic of in-house vs outsouced, and the MindTitan’s opinion on those, spoken by Kristjan Jansons, who is a machine learning expert, CEO and co-founder of MindTitan which has delivered 80 machine learning solutions to the biggest organizations like Elisa, Veon and the Estonian public sector which is often regarded as one of the most digitally advanced countries in the world.

The versatility of the team

To implement the AI project successfully, you will need several specialists, and the work distribution between them is uneven during the project time.

  • Thus, analysts are required from the very beginning of the project for the mapping, to determine what needs to be done from the business perspective and what are the technical demands of the project. Then, their work could be needed more during application development, but usually, their involvement is lower after the initial mapping.
  • Data scientists are required while the model is being actively developed. Afterward, their involvement decreases when a model is set up.
  • Developers are only required at the app development phase to use the benefits of the AI model that was developed after the first usable model version is ready for use.
  • Project manager is required all the time, but relatively less in late project phases.

Additional project members might be required to smoothly bring the project to life. For example, data engineers, DevOps, sysadmins, UI/UX designers etc.

The work distribution between the project roles.
The work distribution between the project roles. The months are for illustration, and time can vary significantly between different projects.

In-house team: budget limitations for high-quality experts

Businesses, especially at the starting point, have financial restrictions in terms of hiring people. There is usually a budget only for a small team.

kristjan jansons

Kristjan Jansons, MindTitan CEO and co-founder:
“When businesses are faced with a data science project, people usually think that a couple of data scientists are sufficient, and they will do everything.

But, to do complex tasks fast and well, there are many more roles to be covered. If a person is building the internal team from scratch or a team does not have any experience working together, then it is very hard to even understand who is good at what.

For example, with the hypothetical perfect project, two data scientists could do most of the heavy lifting, but you still need five other people with complementary skills in different places for a shorter time.

“If you build such a team in-house, it could be time-consuming and inconvenient in terms of finances and efficiency,” Kristjan concludes.

Outsourced team: a wide range of experts for the same price

“Outsourced teams are more versatile than in-house teams for the same price.” (Gergana Krusteva, La Koketa co-founder)

Teams specialized in machine learning projects deal with the full cycle of AI implementation. For example, at MindTitan, turnkey AI solutions could be provided. There are specialists for every segment of the processes, starting from a strategic alignment of AI and business goals, ending with data scientists and developers with broad subject knowledge.

Turnkey AI solution by MindTitan concept
Turnkey AI solution by MindTitan

Learning curve vs practiced team

The learning curve effect suggests that the more times a task has been performed, the less time is required on each subsequent iteration.

In-house team: learning curve, slowing down the process

Internal teams may find it challenging to implement artificial intelligence effectively if they are doing it for the first time. As new tech businesses are usually innovative, the chance that internal specialists have experience in this exact type of project is relatively low.

Outsourced team: experienced professionals at hand

As mentioned above, outsourced teams, working with machine learning projects all the time, have highly qualified specialists on board; thus, they can start implementing the project right away, not spending time learning how to be more efficient. They already know who is good at what and know whom to include to deliver in an efficient manner. This might mean including a specific person with a particular skill set for 1 day or 6 months.

The price-quality rate

As entrepreneurs are interested in the price-quality rate to make their businesses more profitable, a relatively lower hourly price would make them consider hiring internal AI experts.

In-house team: Short-term salaries might seem less expensive, but the expenditure might be not effective in the long run

Contemplating all the financial circumstances (taxes, insurance, training fees, infrastructure, etc.), the price difference becomes not that significant. Plus, when the peculiarities of machine learning technology are considered, hiring an internal team could be unprofitable and even worthless.

Kristjan Jansons, MindTitan CEO and co-founder explains: “In reality, some pauses in your ML project could appear; for example, there could be a long wait for some data collection or labeling, or unfinished work from another person could hold up the whole project, as it is an input for your work. For us, in MindTitan, it is solved easily; we just switch people between projects. With an internal team, only one project in hand, and a limited number of specialists, it could be weeks or sometimes months wasted because people are just waiting for something.

Of course, people can always come to sit at work and do something, but are the salary outlays in such a case justified? Also, what do you do with the in-house team when the core of the AI project is finished, and it enters the maintenance phase? Most AI projects do not need endless and constant effort from the machine learning team. Most AI projects will have diminishing returns once a certain target metric has been achieved.”

Outsourced team: The hourly rates might look expensive but, at the project scale, are more effective

“Outsourced machine learning has allowed us to be pretty lean financially and also get the best people working on their specific expertise.” (Matthew Daley, MAX Sports Health Director of Business Development)

In contrast, an external machine learning team could seem more expensive in terms of hourly price. However, as mentioned above, the process of AI implementation could have gaps and pauses that an outsourced team can handle, optimizing the process with educated data scientists, developers, and other specialists. Also, as the first phase of the machine learning project implementation could require significantly more resources, reinforcing internal specialists with machine learning outsourcing could be beneficial. In the picture below, the core AI development phase is followed by a maintenance phase, where the effort usually decreases with every iteration.

The core AI development phase is followed by a maintenance phase where the effort usually decreases with every iteration.
The work load at different phases of AI development.

Dedication vs breadth

“An in-house team is way more agile and has a much better understanding of the problems to be solved by the AI technology team than an outsourced one.” (Bogdan Predusca, Hyperhuman founder)

This interpretation probably comes from the fact that an internal team has one project to work on, while an outsourced one could have several at a time. Of course, in that sense, we could talk about dedication, but should the process become stuck, limited team numbers mean that it could be difficult to get insights quickly without external help or machine learning consultation.

In-house team: may be dedicated, but less effective when the process is stuck

In contrast, external teams have the luxury of easily obtaining extra thoughts, and extra people in the project.

Outsourced team: dedicated and efficient

Kristjan Jansons, MindTitan CEO and Co-Founder tells: “For example, at MindTitan, we have had many experiences where we might have three or four of our people working on the project, and they are dedicated. At some point, they become kind of stuck. Then, we bring in somebody else for a second opinion, who could say “Why didn’t you try this?” It is very easy to bring in fresh ideas.”

Dedication vs breadth concept
Dedication of the in-house team vs breadth of the outsourced team

Moreover, their dedication alongside proper and systematic management could be strictly fixed in the contract and other legal documents.

atte

Atte Keinänen, Head of Engineering at Fuzu Ltd:
“We have always been data driven, but working with MindTitan was what enabled us to really launch our data science and machine learning work. Together with MindTitan, we can continuously grow and explore new areas where we can use these technologies.

It’s good to know that when we want to take our language models to the next level, we can call on MindTitan’s expertise.

They feel like an extension to our team, rather than just consultants. I know I can reach out and get an answer to my question the very same day”

Sync, communication, and privacy risks

“The knowledge developed by an outsourced team is outside the company.” (Pawel Paczuski, upmedic co-founder)

Unfortunately, sync and privacy issues could appear in outsourced and in-house teams as well: people could quit their jobs and the knowledge might be lost.

The shared problem for internal and external teams

It is the question of the proper knowledge transfer or process recording in general. The answer for both in-house and outsourced teams is easy: set up a documentation process. Yes, it is probably not possible to document everything, but at least the processes should be documented well enough to make taking over the work possible. People seem to remember much better and opt for documenting and recording your work as a good idea when working with an outsourced team, which is a nice side benefit of a machine learning project. However, some big IT companies with significant resources may still rely on long-term, in-house employment, thus retaining key knowledge for internal use.

Another side of the risk is sensitive data that could leak. And, again, this is a shared problem that could be solved by signing an NDA.

The risk of a wrong choice

Finding a reliable partner or employee, whether in-house or outsourced, is a challenge. There is a risk of hiring unfit machine learning specialists, placing the whole project at risk of slowing down.

The shared problem for internal and external teams

“Outsourced machine learning could take more time because you chose the wrong team first.” (Gergana Krusteva, La Koketa co-founder)

It is a challenge to find the right partner for better results. However, it is relatively easier to change an outside team compared to an in-house team replacement because of legal issues and time costs. Hiring and firing processes are much more complicated: anyone who has hired specialists can relate to how much time and human resources it requires just to conduct job interviews. Building an AI team is a separate skill; yes, it is hard to find the right person for the team, but, even if you find the person with the proper skill set quickly, sometimes people in a team just don’t click and can’t work 100% effectively.

It is slightly easier for outsourced teams. First, it requires much less effort to change the team for a better one. Second, it is relatively easier to figure out which company would be a proper partner, just by following our checklist below:

How to recognize the right team for outsourcing machine learning projects:

  1. They have a well-defined process laying out how they work.
  2. What is the partner-to-be going to deliver: is it just an AI model development or a turn-key solution? The second option is better because the project will not get stuck while being deployed and maintained.
  3. They have a track record with similar projects. However, this point could be tricky, because, when it comes to data science projects, business people might not always have the best judgment as to what similar means. For data scientists, some projects actually might be quite similar, while for a businessperson they might look different
  4. The outsourced team has a range of skills and competencies that cover what you are looking for. This point requires extra attention and consultation because first-time tech business leaders don’t always have enough knowledge to recognize the skills they should search for.
  5. They can function both independently and as a part of your team at the same time to create a successful project. However strong the AI team is, if they don’t cooperate with the business side, the project will not work at maximum efficiency.

 

Conclusion

Here is a visual representation of in-house vs. outsourced machine learning teams.

Difference between in-house and outsourced machine learning teams

“As we have heard about hundreds of potential AI use cases by now, we would recommend an outsourced team for at least 90+% of the cases,” Kristjan Jansons, MindTitan CEO and co-founder said.

Even though an internal team for machine learning may seem a more solid solution, they are better options for longer projects of 3+ years.

Whereas outsourcing machine learning solutions perfectly fits complex developments of any duration, even after the development project endpoint, outsourcing companies, with only a few hours a week, can easily maintain what they have helped to build. Machine learning outsourcing is a better choice, both for time and money.

Thus, if it is possible to spend less, why not start doing it?

ai plan execution

Go back