Generative AI for Corporate Planning: How to Improve Market Research, Competitor Analysis, and Management Reporting

Market research AI: getting to a starting point faster

One reason market research eats time is that people dive straight into gathering information.

Take a brief as broad as “look into the domestic B2B SaaS market” — it leaves far too much in scope. Market size, growth rate, customer segments, regulation, technology trends, overseas players, substitute services: the issues worth checking are scattered all over.

Generative AI is well suited to that initial sorting of the issues. The trick is not to ask the AI for an exact market size, but to have it help design the research.

For instance, you might ask the following.

Theme: the growth potential of the domestic B2B SaaS market over the next three years

For this theme, set out the issues the corporate planning department should check before drawing up a business plan.

Output format:
1. Seven key issues
2. The primary sources to verify for each issue
3. The angles to examine under a PEST analysis
4. Points to watch when checking the figures
5. Questions likely to be raised in the management meeting

PEST is a framework for sorting out how the external environment — political, economic, social and technological — bears on the business. Having the AI lay out the PEST angles makes it easier to leave fewer gaps in the research.

What you want from the AI at this stage is not the right number. It is to organise which information to go after, in what order to investigate it, and which angles are easy to overlook.

In other words, the value of market research AI lies less in “producing the answer” than in “designing the research quickly”.

Competitor analysis AI: getting the comparison axes aligned

In competitor analysis, the comparison axes tend to wobble from one person to the next.

One person leans on price, another on features, another still on sales channels or track record. The upshot is that, come the management meeting, it grows hazy what the comparison table is actually meant to help decide.

With generative AI, you can settle the comparison axes for competitor analysis up front.

Design the comparison axes for a competitor analysis under the following conditions.

Subjects:
- Us: a corporate generative-AI training service
- Competitor A: a large training company
- Competitor B: an AI consulting firm
- Competitor C: an e-learning provider

Output:
1. The items to examine through a 3C analysis
2. The items the competitor comparison table should contain
3. Evaluation axes usable for a management decision
4. The sources to verify during the research
5. Items the AI should not assert and a person must confirm

3C is a framework that sorts out the business environment through three lenses: Customer, Company and Competitor. Competitor analysis AI is useful not just for building the comparison table but for working out “what we need to compare in order to reach a decision”.

That said, competitors’ pricing, track record, features and organisational set-up all change. Rather than dropping the AI’s output straight into a document, you need to verify it against official sites, IR materials, press releases and first-hand intelligence from the sales front.

Management-document AI: building the storyline

The hard part of preparing management documents is not lining up the information. It is rearranging it into a flow that makes it easy for senior management to decide.

The familiar problems with management documents run as follows.

• Plenty of research findings, but the conclusion is hard to see
• The business unit’s explanation is pasted in wholesale, with no company-wide perspective
• There are charts and tables, but it is vague what decision they are asking for
• Risks and alternatives are not adequately set out
• There is too little supporting material prepared for questions from the board

Generative AI can be put to work building exactly this kind of storyline.

Using the information below, draft the structure of a document for a management meeting.

Purpose:
To have the board decide on additional investment in new business A

Background:
- The market is trending upward
- Competitors are entering in greater numbers
- Early customers have responded well
- However, there are constraints on the sales and development functions

Output:
1. A storyline for board reporting
2. A slide-structure proposal
3. A title for each slide
4. The issues needed for a management decision
5. Anticipated questions and the direction of the answers
6. The points that need fact-checking

With this approach, you do not expect the AI to finish the document. You have it build the first skeleton, and then a person hones the issues, the figures and the decision points.

It matters to position management-document AI not as something that decides in the corporate planner’s stead, but as an assistant that puts the order of thinking in place.

Learning to design research

The first theme to tackle is research design.

When using market research AI, opening with “tell me about this market” tends to get you platitudes in reply. In training, the first exercise is turning a research theme into a question.

A poor example

Look into the elderly-care market.

A good example

To decide whether to enter the corporate elderly-care support market by 2027, set out the market conditions, customer issues, regulation, competitors and profitability that the corporate planning department should examine.

Once the question becomes concrete, the AI’s output moves closer to something usable in practice.

In training, it pays to cover the following angles.

• Make the purpose of the research clear
• Separate the information needed for a decision from background reference material
• Use PEST, 3C, SWOT and the like as the situation demands
• Distinguish primary sources from secondary ones
• Read AI output on the assumption that it must be fact-checked

Codifying competitor analysis

The next theme to address is competitor analysis.

In competitor analysis, you need to decide which axes to look along before gathering any information. Price, features, customer base, barriers to adoption, brand, distribution network, support set-up — when there are too many comparison axes, the result becomes hard to use for a decision.

In training, the exercise is to have generative AI produce the comparison axes, after which a person narrows them down.

You are in charge of competitor analysis in the corporate planning department.
For the services below, propose the items for a competitor comparison table that can be used for an investment decision in the management meeting.

Conditions:
- At most eight comparison items
- Leave out fine feature differences that do not bear on the management decision
- For each item, explain why it matters
- Finally, set out the primary sources to verify during the research

Through exercises like this, the team learns to use competitor analysis AI not as an “information-gathering tool” but as a “tool for designing the axes of judgement”.

Building the storyline of a management document

In training on management documents, the emphasis falls on storyline design rather than on producing the slides themselves.

What senior management wants to know is, in most cases, the following five things.

1. What is happening
2. Why a decision is needed now
3. What the options are
4. What the merits and risks of each are
5. What the recommended option and next action are

Working from this flow, you have the AI build the document structure.

For the theme below, build the storyline of a document for a management meeting.

Theme:
The feasibility of entering an overseas market

Output:
1. The conclusion to lead with in the management meeting
2. Background and market conditions
3. The business opportunity
4. The principal risks
5. A comparison of options A, B and C
6. The recommended option
7. The matters you want the board to decide
8. Information to relegate to supporting material

Running this exercise lets the team use management-document AI not just to “tidy up the prose” but to “build the flow of the decision”.

Building in fact-checking and review

The single most important thing in corporate planning’s use of AI is fact-checking.

Generative AI can produce natural, persuasive prose. At the same time, it may get figures, sources, proper nouns and the latest information wrong. Established risk-management guidance for generative AI sets out the case for building in the identification, measurement and management of risk whenever it is used.

NIST “Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile”

Training must always cover not only “how to use AI” but “how to doubt its output”.

The items to check are as follows.

• Is the source of the market size clear
• Do the year, region and scope of the figures line up
• Is the competitor information current
• Does it contradict official information
• Are conjecture and fact mixed together
• Is there sufficient grounding to use it for a management decision
• Has any confidential internal information been entered

In AI training, rather than adopting the output as it stands, it matters to build the habit of sorting it into “fact”, “conjecture” and “to be confirmed”.

Before training: take stock of the work and the documents

Before the training, take an inventory of the documents and tasks the corporate planning department actually uses.

For instance, review documents such as these.

• Past management-meeting documents
• Working papers for the mid-term management plan
• Market research reports
• Competitor comparison tables
• Notes from business-unit interviews
• Questions the board raises often
• The business-plan format

From these, separate the parts AI can make more efficient from the parts a person must judge.

There is no need to try to put every task through AI from the outset. The realistic move is to start with work where the effect is easy to see and the risk easy to manage — organising the issues for market research, a first cut at a competitor comparison table, a structural proposal for a management document.

During training: practise on real work themes

During the training, use themes close to your own work rather than generic samples.

For example, exercises along these lines work well.

• Have the AI break down a market-research theme for a new business proposal
• Have the AI build the comparison axes for three competitors
• Have the AI build the storyline of a document for a management meeting
• Have the AI generate the questions the board is likely to raise
• Check the AI output against a fact-checking checklist

In an environment like Kanata, where AI chat and AI summarisation can be used within the same project, it becomes easier to run an exercise end to end — from summarising meeting notes and documents through to organising the issues for a management meeting. Kanata’s materials describe AI chat as a feature for asking questions, seeking advice and requesting drafts, and AI summarisation as a feature that summarises meeting recordings, minutes and documents into a specified format.

After training: turn prompts and checklists into a library

What matters after the training is not letting what was learned end as one person’s skill.

In the corporate planning department, keeping prompts like the following as a shared departmental asset makes the work more reproducible.

• A prompt for organising market-research issues
• A PEST analysis prompt
• A 3C analysis prompt
• A prompt for building a competitor comparison table
• A prompt for building the storyline of a management-meeting document
• A prompt for generating anticipated board questions
• A prompt for fact-checking

When using Kanata, the project library can hold AI settings, prompts and training data, which makes it easier to organise prompts and reference materials specific to the corporate planning department. In its materials, the prompt library is framed as a store for frequently used instructions, and the training-data library as a place for the internal documents you want the AI to reference.

Do not make AI output the basis of a management decision

You must not take prose or tables produced by AI and use them, as they stand, as the basis for a management decision.

AI is effective for forming hypotheses, organising issues, proposing comparison axes and building document structures. But market size, growth rates, competitors’ revenue, pricing, customer numbers and regulatory information must always be verified against primary sources or trustworthy materials.

AI is a “guideline to start thinking from”, not “the grounds for deciding”.

Do not enter confidential information

The corporate planning department sometimes handles unpublished financial information, M&A, personnel matters, new-business plans, investment decisions and customer-level profitability.

When such information is to be entered into an AI, you must follow your organisation’s information-management rules. It is especially important to distinguish information that may be entered into an external service from information to be handled only within an internal-only environment.

The OECD’s AI Principles, updated in 2024, place emphasis on addressing issues such as privacy, safety, information integrity and intellectual property. Corporate planning, too, needs to bear these angles in mind and be clear about the scope of the data it inputs.

In training, alongside how to write prompts, the following rules should be conveyed as a set.

• Do not enter unpublished financial information
• Mask personal names and customer names where necessary
• Check anything touching contracts or NDAs in advance
• When in doubt, do not enter it
• A person must always review before anything goes outside the company

Do not leave AI use to the individual

One reason AI adoption stalls in corporate planning departments is that the manner of use is left to the individual.

One person uses it well; another has no idea how. AI is used in one document; another is done by hand as before. In this state, departmental productivity is unlikely to rise.

For that reason, it is important to establish shared operating rules after the training.

• Which tasks AI is used for
• Which prompts are made the standard
• Which information is never entered
• At what stage a person reviews
• Where the output is stored
• Who updates the prompts

With these rules in place, AI use shifts from the ingenuity of a few people to standard departmental practice.