Is Artificial Intelligence (AI) a risk or a future potential? With an AI software due diligence, investors analyze opportunities and risks of AI-Systems in the target company in an early M&A transaction phase to secure their investment decision.
- Find out whether the right AI model is being used for the use case
- Verify that the AI system has been trained with the right data
- Validate the consistency and preparation of the data
- Check the current prediction quality of the AI models
- Ensure that the software base is state of the art
- Determine the development priorities from Day 1 onward
- Determine if the AI data will continue to be available after carve-out
An AI due diligence with Cape of Good Code
Experience makes the difference
The characteristics of AI software are not obvious. A software can be near the end of its life and still operate ok. But for how long? That's a serious risk to the businesses future. We at Cape of Good Code are software developers and consultants. Together with our partners’ AI-experienced Data Scientists we bring the right specialists to the project. We know the issues that should be explored in the AI-code and in the data. We make sure that any serious flaw in the AI software will be known and can be taken into account in the investment decision.
Quick and efficient
We know that in a due diligence, time is short and budget is tight. Our state-of-the-art DETANGLE analysis suite and our consultants in the project are a well-coordinated team. With clearly identified and reported risks, the investor can make a diligent investment decision.
Feel free to contact me about your personal AI case.
Every AI project is unique and presents different challenges. We always have your company in mind and develop solutions that fit you.
The analysis of the AI software in M&A transactions
Do you want to analyze the target company's AI software or AI system to know the quality of technology before the acquisition? Do you need a realistic assessment of the software-based operational and financial risks? Our AI due diligence provides the answers, possible red flags but also an outlook on possible solutions.
We have analyzed software startups as well as companies with hundreds of developers and millions of lines of code. We can therefore advise the investor during the M&A transaction whether:
- The AI software is ideally suited for the use case
- The AI system and its development is scalable
- Escalating maintenance costs are to be expected
- Red flags exist from a technological and organizational perspective
Contact us and use our know-how from many successful software analysis projects.
Your benefits of our AI due diligence program:
- No technology surprises late in the M&A transaction.
- Certainty that the AI system is the right one and will work
- Realistic assessment of synergies and required investments
- Clear priorities in development from Day 1
- Better negotiation and contract position
High quality of the AI System and its data create high value
Artificial intelligence (AI) brings companies in their business automation and decision-making processes to a whole new level. The technological basis of an AI system is based on software.
The analysis methods for code and architecture quality of a classic software are known and proven (see @Software Due Diligence). With AI software, the additional relevance of the amount of data volume their structure requires a different approach in the software analysis methodology.
An AI system must be trained first with suitable and representative data, before it can go into operations. As this training is key for the prediction quality, the training data and -process has to be part of the risk assessment in the Software Due Diligence. The operational data in turn must be specially prepared and be available in sufficiently large volumes to unleash the potential of the AI system to deliver correct predictions/suggestions/decisions. AI software is still a rather new technology with corresponding unknowns and risks in its use. Very AI-specific experience and knowledge are needed from the consultants in AI-involved M&A transactions.
Cape of Good Code partnered with an established AI consulting firm and has jointly developed a special methodology for a technology-oriented AI due diligence. This involves examining AI-specific risks in addition to software technology challenges such as high technical debt or lack of cloud readiness. Within the tight budget and timeframe of a due diligence, this approach delivers a reliable and holistic assessment of the potential risks and weaknesses of the target company's AI system as well as the data processed in the AI System.
Take a look at our references
How AI software is making new demands on software due diligence
New due diligence criteria to assess the risks of AI software.
(Published in M&A Review 7-8/2022)
What should be considered in the due diligence of an AI software?
Traditional software works by "hard-wired" rules set by the programmers. In M&A transactions, Software Due Diligence has proven effective for evaluating classic software. The data that is processed in that kind of software plays a minor role for the risk assessment. An AI system, on the other hand, has no "hard-wired" rules and must be taught first with extensive and representative process data. With more and more processed data, the system can develop further during operation (machine learning) and will (theoretically) deliver ever more accurate forecasts of trends or events, whose origins, however, are no longer directly traceable.
This is the strength and at the same time the weakness of AI and therefore also a risk factor for an investor: How good was the data with which the AI was trained and how accurate are the results with the data that is operationally available? To evaluate an AI system in due diligence, additional areas of consideration and new criteria are required compared to the classic software due diligence.
What are the possible risks of using AI?
The data essentially determine the quality of the AI system
AI is used to recognize patterns and correlations in huge volumes of rather unstructured data. An AI-System can deliver answers to questions that could not be answered before AI. Thus, both the quality of the data when teaching the AI system and the data in operational use play a key role in the added value or hidden risks of the AI system.
Following risks may arise:
- The AI system has been trained with biased data.
- Operational data is not pre-processed appropriately.
- The data volume is not large enough.
- The data is no longer accessible after the M&A transaction.
A particular danger can occur, if the AI software suddenly and unexpectedly produces incorrect or illogical results, e.g. due to a data scenario that was not previously taken into account. In this case, fallback scenarios must be called up instantly in order to interrupt the AI system, check it and, if necessary, retrain it including the data of any new scenario. This will avoid nonsense results being processed further and cause damage to customers and finally to the company. As there are many examples that this really happens, our team will check this aspect in an AI due diligence.
The right AI system does not fall from the skies
The AI system is an AI algorithm trained with data. Not every algorithm fits every use case. Only with specialized knowledge and experience can it be judged whether the right AI system has been selected for the use case at hand. If a suboptimal AI system has been selected, this becomes noticeable through limited accuracy of the results and a lack of scalability for further growth. Then AI is a risk rather than an opportunity in digitization. Since a key feature of AI is predicting the likelihood of future events or trends occurring, the consequences of incorrect results can pose a very high and not immediately apparent risk to the business.
In our AI due diligence, we use interviews, questionnaires and benchmarks to check whether the AI system fits or obviously does not fit the use case and the investment thesis.
Thereby it has to be checked
- to what extent is the labeled AI-system genuinely an AI/ML algorithm
- were various suitable alternatives considered when choosing the algorithm in order to decide on the best option
- how extensively and sophisticatedly the algorithms available for selection have been validated for their prediction quality.
What requirements must the organization fulfill when using AI?
To use the full potential of AI, a high AI-affinity at management level and in the specialist departments is prerequisite. Open and intensive collaboration between management, the specialist department and IT is the base for the successful introduction of AI systems in specific processes.
Our consultants verify whether AI is being used as a buzzword for M&A marketing or whether it is securely anchored in the company's culture.
Important characteristics of a true AI-open company organization are for example:
- A clear digitization and AI strategy roadmap is in place
- AI metrics are a regular part of management reporting
- AI investments are budgeted separately
- IT has specialized knowledge and capacity
- Business departments are aware of AI goals and opportunities
- Specific Q metrics are in the QA manual for operational monitoring of processes digitized with AI
Are you interested in a AI software due diligence and a personal meeting? I will be happy to help you.
+49 8341 96 111 60
What is the importance of the software technology in an AI system?
Software is the technology that makes AI work. If a good AI system was realized by a poorly designed software, it will show a deteriorating performance and will generate high follow-up costs, no matter how good the results were at the beginning. These risks can be effectively investigated with the DETANGLE Analysis Suite (see Software Due Diligence).
Some of the important questions in the software due diligence are:
- What is the technological future viability of the software?
- Is there adequate code and architecture documentation?
- Can the software be developed and maintained at normal cost?
- How efficient is the software engineering process?
- Are the key developers part of the M&A transaction?
- What are the technological security risks?
When the answers to these questions are right, it is highly unlikely that the software base will limit the success of the acquisition.
How long does an AI software due diligence take?
A Red Flag AI software due diligence, is appropriate at an early stage of the M&A transaction. It can be performed in approximately 6-10 man-days. A AI due diligence takes a bit longer than a due diligence of regular software, because the AI-relevant technology can not be analyzed automatically. Most AI specific details are obtained as expert analysis from interviews and questionnaires.
A full scope AI software due diligence is appropriate when the importance or the valuation of AI is high and the M&A transaction is more advanced. This takes 10-25 man days. It depends on the scope of the software (lines of code), the complexity of the AI system and its data, as well as the requirements and scope of the project reporting and the final report.
In addition to in-depth statements about the possible risks in the AI system and the accompanying processes and structures, it also provides additional information about the effort required to reduce possible critical technical debt in the software base (see also Software Due Diligence). Likewise, risks from dependencies on certain developers or suppliers are also identified and, last but not least, the investor also receives suggestions as to which projects should have which priority from Day 1.
Which investigation modules are used in the AI software due diligence follows the priorities of the investment thesis, the complexity of the target and is concluded from the investor's briefing. The table below gives an overview of the possible areas of investigation in Red Flag or Deep Dive Software Due Diligence.
Red Flag KI Due Diligence
Capture the risk issues of a software in a short time and with manageable effort
Deep Dive KI Due Diligence
Get a detailed picture of the software and recommendations for action for Day 1
|Expert interviews on data preparation and minimising bias in data|
|Statistical analysis of the data before and after processing|
|Investigating the selection and quality of AI system training|
|Investigating the prediction quality of the operational AI system|
|Review of monitoring processes and contingency plans|
|What-if analysis of unforeseen events to review contingency plans|
|AI-Readiness of th organisation|
Are you interested?
Our Latest Knowledge Articles
In our knowledge blog you will constantly find new articles on various topics from software development.