For several years, we have been using artificial intelligence to increase our productivity in updating and publishing sector studies. The exponential evolutions of the latest models have considerably widened the field of possibilities.
I experimented last month with producing a SWOT with GPT3. Here is how this experiment works and my conclusions.
Why do we use an AI to do a SWOT analysis?
The SWOT analysis – Strengths, Weaknesses, Opportunities and Threats – is a very classic and well-documented reading grid in strategy. It is therefore a model widely used and exploited by AI.
SWOT analysis is also a simple reference approach for a user to understand.
How the company’s SWOT is generated with an AI
The SWOT is generated in several stages:
In just a few days, more than 250 very varied SWOTs were created on Businesscoot. Many users have chosen SWOTer their company. They were SMEs, start-ups and many international groups. The markets involved range from the global automotive market to micro-niches such as the organic aromatic herbs market.
«The variety of results allowed me to have a complete vision of the strengths of this model as well as its flaws» says one of the users.
To measure the quality of the results, I considered that a Strength, Weakness or other proposal was successful if it was «human like», as if it had been carried out by a human after analysis of the company and the sector of evolution concerned.
1- AI (GPT3) does not do EVERYTHING:
The SWOT is produced in 3 stages:
- From the website sent by the user, I define the business of the company;
- From that of the 2 main competitors, I identify the company’s market;
- With these elements, I produce a SWOT on the company.
and here is the performance of AI:
- In 98% of cases, AI is able to understand and summarize the activity of the company, whether it is a small business or a multinational;
- In 92% of cases, AI correctly identifies the market, even for a niche market such as «the market of organic herbs».
- On the other hand, the production of the SWOT is more laborious. Here is the percentage of success to identify the strengths, weaknesses, opportunities and threats of a company: Strengths: 92%, Weaknesses: 37%, Opportunities: 35%, Threats: 35%
successful = Human like result
2 – Results vary based on proposed data
- Strength: 92%
I feed my AI with the content of the company’s website that I «SWOT». AI translates (rather well) in «Strengths» the information of the site.
But it will not challenge these Forces. When a company describes itself with «exceptional customer service» one of its «Strengths» will be: customer service (surely true, but which does not have «exceptional» customer service:p).
- Weaknesses: 37%, Opportunity: 35%, Threats: 35%
On the other hand, AI must manage to find Weaknesses, Opportunities and Threats, most likely from the content of tens of thousands of SWOT analyses. The results are already less relevant.
- In 78%, competition appears in Threats.
I send my AI information about 2 competitors. AI understands that I identify competition as an important element and seeks to insert it into the SWOT.
The data sent to AI will strongly influence the results of my SWOT.
3 – SWOT standardization
Out of a sample of 30 SWOT products, in 76% of cases, «notoriety» is an «Opportunity» for the company.
- Is that fair?
On the Businesscoot SWOT, notoriety is an Opportunity (which is true). However, this conclusion is not directly related to the understanding of Businesscoot’s positioning but to a statistic: when the company is not known, the improvement of notoriety is often an opportunity.
AI does not look for the weaknesses of a company but the most likely weaknesses for similar companies.
This allows a reminder: an AI is built on probability statistics.
The most common answer is the best, which may be in contradiction with a company-specific SWOT.
Here are other examples that are often repeated in the 250 SWOTs I studied :
Most common opportunities: visibility, social networks, awareness, product line, partnership, technology (if market is technology related).
Most common threats: competition, price, inflation.
Some examples and improvements to consider in the next release
- For the SWOT of a market leader, informing competitors strongly guides the result and standardizes responses. This point is not clearly shown at the moment, however informing direct competitors improves the quality of the result, especially if the company is known and the model probably consulted SWOTs about the company.
- Identification of the Forces: the model takes the strengths presented by the company and will not put it in perspective. E.g. “Exceptional customer service”.
- SWOT d’Eiffage: «Valorisation de l’humain» is a strength of the company. This is the case in which the model has identically reproduced an element of the description of the company as such. In the threats to Eiffage, the model cites rising material prices and regulation. It is very likely that the model took elements of a SWOT existing on Eiffage.
- The decoration market: threats: fashion impact: By properly targeting the market, the model identifies a threat that is quite relevant for the SWOTée company.
- Technology as Opportunity: in many market cases, technology is an opportunity which is often correct. However, the model is sometimes off topic, as in a SWOT for a company in the creative leisure market. On the other hand, international development has been identified as an opportunity, a response adapted to this market.
- Opportunities in the dog and cat food market: marketing. This opportunity seems relevant to me and comes from a correct identification of the market, more than from the company SWOTée, which is an SME on which the model probably has no data.
- Opportunity: Market growth. Once again, it is thanks to a correct identification of the market on which the company is positioned that the model manages to deduce that market growth is an opportunity for the company.
- The furniture and decoration market (SWOT for IKA)=> weakness: lack of notoriety. I do not understand how he could be so wrong about an industry leader.
- The furniture and decoration market: the market is correctly identified and the opportunities are coherent.
- The market for office and school supplies: the threats are particularly specific: price increases and imports.
- Lack of coherence between strengths and weaknesses: the market for organic herbs: the local dimension is taken over and in strengths, and in weaknesses.
- 2 SWOTs are different for the same company for «the market of cold-pressed organic vegetable juices»: The demand for organic products does not appear in the second SWOT while the market identified by the model is the same.
All these elements highlight the strengths of the model and the weaknesses and improvements to be made in the next version.