Gains:
- Being able to distinguish where in the consultancy chain (research, analysis, synthesis, suggestion, presentation) artificial intelligence saves real time and where judgment and responsibility are left to humans, depending on the task risk level.
- Ability to apply a discipline that verifies each artificial intelligence output through the steps of connecting it to the source, recalculating it and passing it through a logic filter.
- Ability to establish a working order that protects customer data within the scope of KVKK and confidentiality agreements, chooses safe vehicles and observes ethical boundaries.
Management consultancy is the profession of helping to solve a difficult problem faced by an institution (company, public institution or foundation) from an external, impartial and structured perspective. Advisor; It collects a lot of information in a short time on a subject such as growth, cost, organization, strategy or operation, synthesizes it (transforms the parts into a meaningful whole) and offers a clear recommendation to the decision maker. The essence of the matter is not knowledge; It is the insight that turns into a decision (insight, that is, the meaning extracted from raw data that motivates action).
Artificial intelligence (here, especially large language models, that is, text generating systems such as ChatGPT, Claude, Gemini) is creating a radical change in this profession. But if used incorrectly, it can cause harm just as quickly. Throughout this module, we will position AI like a senior assistant: a blazing fast researcher, a tireless drafter, a patient thinking partner. But never the signing partner. The signature, that is, the responsibility for the advice given to the customer, always belongs to the human.
Where artificial intelligence comes in handy in a consultant's day
A classic consulting project roughly goes through the following steps: problem definition, data and information collection, analysis, synthesis, proposal development, presentation and implementation support. Artificial intelligence touches every link of this chain, but it does not have the same authority in every link.
- Research and scan: Summarizing a 40-page industry report in 3 minutes, breaking down 15 competitors' websites into a structured spreadsheet, listing the assumptions needed to estimate the size of a market. Artificial intelligence is very powerful here.
- Analysis and modeling: Interpreting a trend in an Excel spreadsheet, sorting open-ended responses from a survey into themes, finding the bottleneck of a process. It is strong here, but it is necessary to re-check the numbers.
- Synthesis and recommendation: "Which market should this company enter?" The answer to a question like. Here, artificial intelligence produces drafts; The consultant establishes the decision and justification.
- Relationship and judgment: How to break bad news to the client's CEO, what suggestion to highlight, what risk to take. This is entirely human space.
Seeing this distinction in a table makes things clear.
advisory role
The role of artificial intelligence
Whose word is final?
Report/news summary
High: fast distillation
Consultant (confirms source)
Set up a competitor table
High: configuration
Advisor (verifies each cell)
Market size calculation
Medium: draft + account
Advisor (cross-check)
Strategic suggestion
Low: generating ideas
Advisor (decision and justification)
Presentation to CEO, relationship
Too low: proof
completely human
Decision in the regulated area
Draft only
Competent expert + human
Tip: View AI as a “thinking accelerator,” not an “answer machine.” The best consultants use it to quickly test their own hypotheses (not yet proven guesses), not to do it for themselves.
Why verification is non-negotiable
The most dangerous feature of large language models is this: they appear confident even when they are not. They may invent the turnover of a competitor, the growth rate of a market or the name of a resource; This is called hallucination, fabricated information that the model presents as real. In consulting, this is fatal because the client makes million-dollar decisions based on your numbers.
That's why the immutable rule of this module is this: No number, no source, no claim coming out of artificial intelligence enters the slide without being verified. Verification is three steps:
- Link it to the source: Ask where each important number comes from and open that source yourself and see.
- Recalculate: Repeat each calculation, such as market size, ratio, percentage, etc. manually or in Excel.
- Logical filter: “Is this number reasonable based on my industry knowledge?” ask. There cannot be an 8 million player in a 5 million market.
Security and compliance - critical areas: it's up to the person to decide
Consultants often work in regulated industries such as healthcare, finance, energy and pharmaceuticals. In these areas, artificial intelligence output does not replace competent expert approval. Capital adequacy recommendation for a bank, treatment protocol efficiency for a hospital chain, licensing strategy for a pharmaceutical company... The final say in all of these; It is subject to the approval of the relevant lawyer, actuary, physician or industry expert. Artificial intelligence prepares the draft, expert verifies it, human decision maker signs it. “The AI said so” is not a defence.
Privacy: customer data is sacred
Consultants see the client's most sensitive data: turnover, salaries, contracts, strategy. Pasting this data into a random AI tool is a violation of most contracts and KVKK (Personal Data Protection Law, the Turkish law that prohibits unauthorized processing of personal information). Corporate tools that guarantee data retention should be used; If not available, the data should be anonymized (names and distinctive numbers are masked and given to the model).
Caution: A customer agreement usually says that "data cannot be shared with third parties." Data pasted into a publicly available AI tool may be considered “shared” to that tool. If in doubt, anonymize the data or do not upload it at all.
three mini cases
Case 1 — Fictitious market number. A consultant asks artificial intelligence, "How big is the Turkish pet food market?" he asks; The model says "approximately 12 billion TL". The consultant puts it in the proposal without verifying it. During the presentation, the customer's marketing director smiles: "We alone make a turnover of 9 billion." The number was made up; The offer loses credibility. The right way was to establish the market size from the base (how many households, how many pets, average spend) and cross-check it with three separate sources.
Case 2 — Breach of confidentiality. A team member loads the payroll for the client's 2,400 employees into a free web tool and says "summarize this data." Data runs the risk of being mixed into the tool's training pool. The confidentiality clause in the project contract was violated; If the customer finds out, it's over. The right way was to use corporate approved tools or replace employee IDs with code numbers.
Case 3 — Good handling. A consultant gives the 80-page annual report of 6 competitors to the artificial intelligence and says, "Make a table of growth strategy, weakness and price position for each competitor, rely only on the report I give." 2 hours of work is reduced to 15 minutes. Then it checks each cell against the source report. Saves time, doesn't make mistakes. This is the targeted balance.
Weak prompt / Strong prompt
A prompt is the instruction you give to the AI; Its quality directly determines the quality of the output.
Weak prompt:
Give information about the Turkish logistics industry.
This prompt produces a mass of contextless, roleless, and unverifiable text; The model fills in the gaps with fitting.
Powerful prompt:
Your role: senior research analyst working for a management consultancy. Task: Structure the Turkish road freight market for an investor brief. Scope: market size range, growth drivers, top 5 players, regulatory risks. Rules:- Rely only on the sources I pasted below; Write down what you don't know as "not in the source". - Next to each objective claim, state in parentheses the source from which it comes. - Don't make up numbers; label "needs verification" where you are unsure. Format: header list + 3 lines "most critical uncertainties" at the end. [SOURCES: ...]
The second prompt gives the role, task, resource, boundary, and format; narrows the field of hallucination.
Four principles of using artificial intelligence safely
Keep the following four principles on your desk like a card; Each module is built on these.
1) GIVE CONTEXT: role + task + source + target audience + format + boundaries.2) CONNECT TO SOURCE: every number and claim must be based on a traceable source.3) RECALCULATE: never entrust the arithmetic to the model, check.4) HUMAN SIGNATURES: critical decision and expert approval in the regulated field is a must.
A role description template you can use when starting each new conversation improves the quality of the output from the start:
You are a meticulous senior analyst working for a management consultancy. Your immutable rules are: - Rely only on the resources given to you; If you do not know, say "it is not in the source". - Do not make up any numbers, names or sources; Mark "must be verified" if you are not sure. - Write the source next to each objective claim. - Do not claim causality; show only patterns and assumptions. Follow these rules throughout this conversation. If you understand, say "I'm ready" and wait for the task.
When you have to share confidential data, impose anonymization as a rule on the model:
I'll give you a data set. Apply this anonymization before processing:- Replace contact names with K1, K2...- Replace company/brand names with K1, K2...- Generalize unique details that can indirectly identify the person (like "only female manager"). Then just work on the anonymized version and show me that version too.
There's also a simple template to keep track of verification:
Output: [what I produced]Source: [which document/page/URL]Verified by: [who, when]Status: [confirmed / unverified / awaiting expert approval]
Common mistakes
- Mistaking the model for authority. Fluent and confident language is not a guarantee of accuracy; The hallucination is also fluid.
- Writing prompts without context. Prompts such as “analyze” produce exaggerated and unverifiable text; Give role, resources and boundaries.
- Uploading confidential data to an uncontrolled vehicle. Violation of KVKK and confidentiality agreement ends the project and reputation.
- Bypassing expert approval in the regulated field. Health/financial outcome is draft; The final say lies with the expert.
- Not keeping a verification trail. "Where did this number come from?" If you cannot answer the question, the issue is not ready for publication.
In summary
Artificial intelligence is a powerful assistant that accelerates every step in management consulting, from research to drafting, from analysis to presentation. But its authority varies depending on the risk level of the mission: high in screening, very low in decision and regulation. The constant rule is verification; No number, source or claim is included on the slide without verification. Customer data is protected by KVKK and confidentiality agreements; The ultimate responsibility and signature always belongs to the human. Throughout this module we will learn each technique along with this security framework.
Application task
Choose your own (real or imagined) consulting project. Break the project's typical workflow into five steps, similar to the table above. For each step, fill in the AI's role (high/medium/low) and the "whose say is final" column. Then, in one step, create a sample record using the verification trail template: which output, which source, who verified, what is the status.
checklist
- [ ] I positioned AI as an "assistant", not a "decision maker".
- [ ] I determined the risk level of each mission and who had the final say.
- [ ] I planned three-step verification (source, recalculation, logic) for each critical output.
- [ ] I evaluated customer data in terms of KVKK and privacy; I anonymized it if necessary.
- [ ] I added expert approval in regulated areas to the process.
- [ ] I wrote my prompts with role, resource, boundary and format.
- [ ] I have committed to using the verification trail template.