AI Strengths and Limits
By the Syncfusion DevTutorials team · Last updated
Learn what AI chat tools do well and where they go wrong. Watch the AI invent a policy and get a sum wrong, sort 12 tasks, and learn how to stop it making things up.
Already use AI at work? You may be able to skip this lesson. Take the short check below.
Ask It About Your Job
In the last lesson, the AI guessed Priya's dates. Now let's see how far that guessing can go.
The box below asks the AI to write your weekly report. It has never met you. It does not know your role, your projects, or your customers.
Press Send and see what it does.
The AI did not say, "I don't know your job." Instead, it created a report with projects, numbers, and achievements.
The problem? None of those details came from you.
In simple words: AI would often rather guess than admit it does not know. It is your job to spot those guesses before they become mistakes.
TL;DR
- AI is very good at working with words: drafting emails, rewriting messages, summarising documents, brainstorming ideas, and organising information.
- AI is weak at facts it was not given, including company information, recent events, customer details, and exact numbers.
- AI can make mistakes in calculations and still sound completely confident.
- A hallucination happens when AI creates information and presents it as if it were true.
- The best way to reduce mistakes is to provide the facts, ask for working where needed, and tell the AI what to do when it does not know the answer.
A Real-World Example
Imagine a well-read colleague who has read thousands of books, reports, and emails. They write beautifully, work quickly, and never get tired.
But they joined your company this morning.
They do not know your customers, your projects, your prices, or your policies.
Ask them to improve a rough email, summarize meeting notes, or rewrite a customer message, and they will probably do a great job.
Ask them for your company's sales figures or internal policies, and they may confidently make something up.
An AI chat tool is very similar. Use it for the jobs it is good at, and be careful where it is weak.
What You Will Learn
In this lesson, you will learn:
- What AI is good at.
- What AI struggles with and why.
- What a hallucination is.
- Why important calculations should always be checked.
- How to reduce the chances of AI inventing information.
- How to work with company documents more safely.
What AI Is Great At
Working With Words
AI learned from a huge amount of written text. Because of that, it is very good at tasks that involve language.
Many people use AI at work every day for tasks such as:
- Writing emails
- Summarising documents
- Drafting announcements
- Preparing meeting notes
- Brainstorming ideas
- Rewriting content for different audiences
Here are some of the jobs AI generally does well:
- Drafting: Creating a first version of an email, announcement, article, or message.
- Rewording: Making content more professional, friendly, concise, or clear.
- Summarising: Pulling key points from longer text.
- Reorganising: Turning rough notes into structured lists or tables.
- Brainstorming: Generating ideas quickly.
- Explaining: Making difficult topics easier to understand.
In all of these situations, the information comes from you. AI helps shape and improve it.
Try a Real Strength
The box below contains a rough and slightly rude message.
Press Send and ask the AI to rewrite it so it sounds professional and polite.
Notice what happened.
The facts stayed the same. Only the wording changed.
This is exactly the kind of task where AI can save time.
What AI Is Weak At
Facts It Was Not Given
AI does not automatically know information about your company.
That includes:
- Internal policies
- Customer information
- Project details
- Team plans
- Prices
- Procedures
It may also be missing recent information, such as:
- Today's news
- Recent exchange rates
- Current stock prices
- New company announcements
When AI does not have the information it needs, it may still generate an answer that sounds accurate.
Hallucinations
A hallucination happens when AI does not know the answer but creates one anyway and presents it as if it were true.
It is not trying to deceive anyone.
It is simply predicting what words are likely to come next.
The problem is that a hallucination can look exactly like a correct answer.
Watch Dev from HR ask for a summary of Kestrel's leave policy without providing the actual policy.
The AI generated holiday allowances, carry-over rules, and sick leave details that were never provided.
If Dev shared this information with employees, people could make decisions based on policies that do not actually exist.
A hallucination often sounds as confident as a real fact. Always verify names, dates, numbers, policies, and rules against a trusted source.
Numbers and Calculations
AI can write text extremely well, but it is not a replacement for a calculator, spreadsheet, finance system, or reporting tool.
Sometimes AI gets calculations wrong while sounding completely certain.
Lakshmi in Finance needs a total for five invoices.
Send her message, then ask the AI:
"Show your working step by step."
The working allows you to see each step and spot mistakes more easily.
In simple words: If the number matters, verify it. For important calculations, use spreadsheets, calculators, or approved business systems.
Sort the Tasks
Now test your instincts.
For each task, decide whether AI is great at it or whether you should be careful with it.
Great at, or careful with?
Look for the pattern.
The strongest tasks involve working with information you already have.
The risky tasks involve facts, calculations, or information the AI was never given.
Stop It From Inventing Information
Give It the Real Text
The best way to reduce hallucinations is to provide the real information.
If the AI has the actual document, policy, or procedure in front of it, it has less need to guess.
However, there is one more important step.
Tell the AI to stay within the information provided.
For example:
- "Use only the text below."
- "Answer only from this document."
- "Do not add information that is not included."
Now help Dev.
Open What you know, provide the real policy information, and ask for a summary that uses only those rules.
With the real information and a clear boundary, the AI is much less likely to invent additional rules.
Easy way to remember: give the facts, then set the limits. "Use only the text I provided. If the answer is not there, say so."
Tell It What To Do When It Doesn't Know
Sometimes the answer simply is not available.
If you do not tell the AI how to handle that situation, it may try to fill the gap itself.
Helpful instructions include:
- "If the policy does not say, tell me."
- "If the answer is not in the document, say so."
- "If you are unsure, do not guess."
- "Only answer from the text above."
You will practise this skill in Exercise 3.
The practice AI used in these lessons is much smaller than many workplace AI tools. It may make mistakes more often. However, the habits you learn here apply to every AI tool.
Common mistakes
| Mistake | What Happens | How To Fix It |
|---|---|---|
| Asking about company information without providing facts | The AI invents policies, prices, or details | Provide the real information |
| Trusting calculations without checking | Incorrect numbers enter reports or payments | Verify calculations independently |
| Sharing a document without setting limits | The AI adds extra rules or information | Tell it to use only the provided document |
| Failing to say what should happen when information is missing | The AI fills gaps with guesses | Tell it to say "I don't know" or "the document does not say" |
| Assuming confidence means accuracy | Hallucinations pass unnoticed | Verify important facts before using them |
Exercise
Complete each task below and then press Check.
What to Remember
- AI works best as a writing and thinking assistant.
- It is great at drafting, rewriting, summarising, organising information, and brainstorming.
- AI is not a reliable source for facts it was never given.
- A hallucination is information that AI invents and presents as true.
- Provide real documents and facts whenever accuracy matters.
- Ask for step-by-step working when calculations matter.
- Tell the AI what to do if the answer is not available.
- Always verify important names, dates, rules, and numbers before using them.
Key Takeaway
Use AI to work with information you provide, not as a source of facts it was never given.
The more accurate information you give the AI, and the clearer your instructions are, the more useful and reliable the answer will be.
Next: The CRAFT Method