How To Make Your AI Project Successful – Business Leaders Must Know

Business Leaders’ Questions

Hi! My name is Trung Dinh, I am the CEO of Aigenexpert, a Data Science, AI Consulting and Development Services company. Over the last few years, I have worked on multiple AI projects in different industries, from healthcare to finance to e-commerce. I have helped quite a number of executives, business leaders, CTOs, VP of different companies, of different sizes in refining their ideas, planning their AI initiatives, and implementing them. There are a few questions that tend to pop up very often, so I hope that you might find the answers to some of your questions in this blog post.

The overall theme I see from those questions, coming from those decision-makers and business leaders, is that: “Alright, we have roadmap and plan, we have a clear definition of success and outcomes, we shall make an investment. How to avoid AI projects fail, and ensure AI implementation success? What is the most important contributing factor that can lead to a successful machine learning project? Where do I find the AI talent? How do I choose an AI vendor? How to avoid wasting tremendous amount of money and time and not realizing until 1-2 years down the line?” And so on.

As you have the plan and goal in place, what’s next then? This is the planning and architecting phase that causes 85% of AI/ML projects to fail after the implementation period. These are the common problems and how to solve them.

Problem 1: Data strategy

The number one factor that negatively impacts all AI projects is the lack of data strategy, the lack of data quality, and data volume. It means that the organization is not there yet for successful AI transformation. They must invest in data first, or alternatively build a POC and subsequently learn about what they actually have and will need, and expand from there if the POC is successful.

But sometimes, such data can be acquired and available, or with certain processes to gather them, or they can come from partnerships. If that’s the case, good news, you can start, immediately. A common misunderstanding many people assume is that if they have lots of data, they must be ready. But such a large volume of data would be worth nothing if they inherently have no predictive quality, actionable insights that can provide any valuable outcomes. So that’s why the lean methodology works in this case, a POC will help you learn what you need to learn. And if the POC is successful, you can continue to build a production-level application from it.

Problem 2: Technical Skills

Assume that you have had the data covered, with the right data governance & infrastructure. Then the second most influential factor is the lack of skills, you need a variety of skills and solutions to make it possible. We have Technical skills and Business skills.

For technical skills, you will need actually competent data science professionals. The difference between a data science/ AI project vs a traditional software development project is the ambiguity and uncertainty of an AI project. In traditional software development, you might add one feature after another, layer after layer, and build upon what you have built up to that point. But in AI, if you have to change the approach that means … sometimes almost everything needs to be rebuilt again, and you might have to discard what you have built up to that point.

The difference between the ones who will deliver and those who do not is very hard to recognize in the interviewing process. Inexperienced or good enough “data scientists, data / AI engineers” often lead to false starts, bad models that look good, and lots of wasted time and money… Because they might not be familiar enough with all the life cycle of a data science project: from data engineering, data processing, building ML models, MLops, building and deploying production-level applications. Many actual data scientists at organizations I saw focus only on building models, but are not familiar enough with the rest. Building a model is only 20% of the whole process… it’s no way near what you have to do for real-life AI/data science applications. You might have to have a data infrastructure in place, gather data, label it, process it, prepare the messy data for actionable insights, then you need to build the AI model, then deploy the model to production, update it, maintain it, monitor it, MLops that kind of things.

How do I hire AI Developers / Data Scientists / ML Engineers etc.?

My advice when hiring is you must be able to assess the true level of expertise and competency of candidates, beyond just degrees and qualifications.  The thing is the technology is changing so fast that things get outdated in 1 year, even months. So those who can adapt fast, with the mindset of constantly keeping themselves updated tend to outperform by a huge margin. The point is that it’s very challenging to hire the right people that WILL deliver. Otherwise, you will not find out until after at least 9-12 months, It could mean up to hundreds of thousands of dollars and tons of time wasted. Usually, it takes 6 months and above to hire one in most developed countries. And that includes the risks that your recruitment process is not optimal, you might have the B-players to interview people and the B-players hire the C-players and the C-players hire the D-players and it all goes downhill from there. And then the fault choice is based on the degrees and past experiences of the candidates… and you might end up with an average person that was just a cog in the wheel in their last projects, with shiny qualifications but just cannot and will not deliver. So make sure you can invest right and hire right, or find a good AI partner to help.

Personally, when we hire people for my team, we don’t emphasize on asking them textbook and theoretical data science questions, we ask them about a hypothetical challenge and see how they would solve it, what range of approaches they might consider, how they would conceptually tackle the challenges and get the information. We ask about a project they have done, ask them about the inefficiencies and the well-known problems with that AI problem? How did they solve them, or not at all…? We ask them to explain their approach, ask more follow-up advanced questions, grill them to the point it takes a long long time for them to think and answer, that’s where we know what their current understanding limit is, where they are in terms of their skills.

How much does it cost to build an AI system?

Bigger projects might require a whole data science team that covers everything from data engineering, AI models, data analysis, and MLops (you might need data engineers, data scientists, data architects, ML engineers, data analysts …). But other projects might require just 1 or 2 data scientists/ML engineers to build an ML module and integrate it to your core, existing software. So the costs depend on your specific problem and the partner you work with, it might not be as expensive as you thought, or as the big consulting firm’s reports make them out to be. You might pay 3-5 times more expensive price and receive equal results. Or you may choose a very low-cost partner but the project fails or the quality is unacceptable. Or you might have the budget and go for in-house only. To solve such challenges that companies face,  at Aigenexpert we have reasonable pricing while guaranteeing exceptional quality of work. We are confident that we are one of the best value-for-money AI companies out there. If you want to know how much your AI project would need and cost approximately, contact me, I will leave my information in the description below. Now. Let’s move on. Still about skills, but this time, business skills!

Problem 3: Business skills

So, it is not just the lack of technical skills of your data scientists. It’s also the lack of business skills in the Data Science team. The gap between understanding how AI technology can impact the bottom-line and benefit your customers and your companies (which requires the business acumen), and the technical know-how of what is feasible and how to build such solution (which requires the technical knowledge), usually lead to a conceptual product and roadmap which is destined to fail in the first place, or a wrong end-product which either does not meet the business requirements or simply inferior and unusable. Capable data scientists, AI people are super expensive and hard to find. And those who possess the business understanding, and people management skills to manage stakeholders, engineers and product people are much rarer. Why is that important? What I don’t see often is someone with such combination of skills being there steering the way and guiding the tech team, the stakeholders, the product people on a common path. That person should be able to explain highly technical models to an audience unfamiliar with the math and statistics knowledge gained over years of study and experience, and connect that to business benefits driven by the results. That person should help all stakeholders with their questions and concerns and bring the group to a mutual consensus. The ones who don’t possess such skills will not be able to justify their work, or they might explain things in a way that is hard to understand, and leave different stakeholders with fractional knowledge and uncertainty and indecision. That could bring the project to paralysis. In some organizations, one key stakeholder’s objection could cause a project to end before it gets the results.

Problem 4: Projects are too complex

Enterprises see that AI projects are extremely expensive. Such costs create a tendency or expectation to have such a hyper-ambitious goal that, this project will completely transform the company and solve great problems with an astronomical return on investment. That usually leads to unrealistic objectives and moonshot projects that usually fail.

At Aigenexpert, we recommend the old-age SMART goal, starting with a single, achievable solution to a specific problem. Successfully build it, get quantifiable achievements. That will create the blueprint for AI adoption at your organization. You can expand from there to more and more successful AI initiatives.

Problem 5: Senior management

The final problem, senior management doesn’t want to invest or stop way too early. The high cost, the high risk, the difficulty of hiring Data Science, AI people, the lack of that expertise in-house; and sometimes the bad experience of working with other company that gives false promise can lead to a risk-averse attitude. My take on this is: if you are a very large organization in which affording a huge data science/AI team is not a problem, or you are an AI-focused product company and have a quite comfortable budget or funding, you can build your own in-house team, invest in it, but take on a few consulting projects with an external partner from time to time just to get the additional quality inspection, perspectives, and insights to improve. But if you are someone with less budget, can’t afford the time it takes to build a full-fledged data science/AI team, and are concerned about the extremely high risks and costs of doing so, then taking on another partner is your best bet. You MUST choose wisely and be able to choose the right partner. DON’T go for the cheap vendor, it’s much more expensive and costly to work with the wrong ones. It could jeopardize your whole business. In Data Science and AI, the project success depends directly and greatly on the intelligence level and skills of the selected few engineers, much more than in a traditional software engineering project. So good enough people are not good at all, neither are cheap-pricing proposals, avoid that.

So, these are the 5 main problems and my advice on how you could avoid and overcome them. If you have a clear plan for your AI transformation roadmap and your AI projects. You know what AI and data science solution you need to build. And you want to find a trustable partner, a company that has the expertise, integrity, and transparency whom you can rely on to drive your AI initiatives to success. I strongly, strongly suggest you contact us today, or even better, schedule a call with me and I will be your thinking partner. We’ll make sure to bring the cavalry to your aid!

See you soon!

About the author

Table of Contents

Get more awesome content from us

Subscribe and get regular updates.
No spam or BS. Just great, insightful articles!

Estimate the cost of your AI project


Send your request to get an estimate for your project, together with Aigenexpert’s actual rates.