Industry8 min read

How Can IT Support Desks Use AI?

On this page

IT support desks can use AI to summarise long tickets, turn fixes into knowledge-base articles, write plain-English replies to staff and build step-by-step troubleshooting checklists. The support officer still checks every step and decides what to run. Passwords, API keys, logs with personal data and system secrets never go into a public AI tool.

TL;DR:

  • Start with writing work: ticket summaries, replies to staff and knowledge-base articles.
  • Mask every ticket first. Usernames, emails, phone numbers and asset tags come out.
  • Never paste passwords, keys, config files or raw logs into a public AI tool.
  • Test any command or script AI suggests on a test machine before it touches a live system.
  • Kenya's national CERT warns that attackers now use AI to write convincing phishing. Your desk is the first line.

This guide is for first-line and second-line support staff, service desk team leaders and the one-person IT department in a small Kenyan organisation. It is general information, not security or legal advice.

The four Cs, before you start

Every prompt below follows the four Cs we teach at FuKazee: Clarity (what you want), Context (who it is for and what the tool cannot guess), Constraints (length, format and what to leave out) and Customisation (your organisation's tone and tools). Square brackets are for your own masked details. For more general prompts, see our ChatGPT prompts for work.

Six ways an IT support desk can use AI today

1. Summarise a long ticket for handover

A ticket that has bounced between three people for a week is hard to pick up. AI can turn the thread into a short note.

Clarity: Summarise this support ticket for the engineer taking it over. Context: it has passed through three people and the user is frustrated. Constraints: under 120 words. Use these headings: the problem, what has been tried, what was promised to the user, what is still unknown. Do not guess a cause. Customisation: plain words, no blame. TICKET: [paste the thread with names, emails, phone numbers, asset tags and IP addresses replaced by labels such as USER-1 and LAPTOP-1]

Check: read the original thread once. AI summaries sometimes drop the one detail that matters, such as an error code or a promise made to the user.

2. Turn a fix into a knowledge-base article

The best knowledge base is written straight after a problem is solved. AI makes that quick.

Write a knowledge-base article from my notes below on fixing [Outlook not syncing on a staff laptop]. Two versions: one for staff to follow themselves, in numbered steps with no jargon, and one for IT staff with the technical detail. Mark any step that needs admin rights as [IT ONLY]. Do not add steps that are not in my notes. NOTES: [paste your masked notes]

Check: follow the staff version yourself on a test machine. If a step confuses you, it will confuse the reader. Our guide on how to write SOPs for AI explains how to keep these documents short and usable.

3. Plain-English replies to staff

Most users do not want to know what DNS is. They want to know when they can work again.

Rewrite my technical update below as a reply to a member of staff who is not technical. Under 80 words. Say what happened, what we are doing and when they should try again. Do not promise a time I have not given. Warm and calm, no jargon. UPDATE: [paste]

Check: every time, date and promise matches what you can actually deliver.

4. Troubleshooting checklists for first-line staff

Write a first-line troubleshooting checklist for "[staff cannot print to the shared office printer]". Order the steps from quickest and safest to slowest. For each step, say what result means "move to the next step" and what result means "escalate to second line". Mark any step that changes settings as [CHECK OUR POLICY]. Under one page.

Check: test the checklist against your real printers, network and policies. AI writes a generic list; your environment is never generic.

5. Understand an error message or a vendor's documentation

AI can explain an error code or a long vendor article in plain words, and suggest where to look next.

Explain this error message in plain English and list the three most likely causes, from most to least likely. Point me to the kind of official vendor documentation I should check for each. Do not give me commands to run yet. ERROR: [paste the error text only, with usernames, server names and paths removed]

Check: confirm the cause in the vendor's official documentation. If AI later suggests a command or script, read it line by line and run it on a test machine first. AI can invent switches and settings that do not exist. Our post on AI coding tools for software developers covers how to review AI-written code.

6. Phishing warnings staff will actually read

Write a short WhatsApp-style notice and a slightly longer email for all staff about a phishing email pretending to be from [our payroll team]. Explain three signs to look for, what to do if they clicked, and who to call. Under 100 words for the notice. Calm, not scary. Do not include any link.

Check: the reporting steps and contact are your real ones. Never paste a suspicious email into a public tool with staff names or working links still inside it.

What an IT support desk must not do with AI

  1. No passwords, API keys, tokens or recovery codes. Not even "just to check the format". Treat a pasted secret as leaked and change it.
  2. No raw logs or config files. They often hold usernames, email addresses, internal server names and settings an attacker would love. Strip them to the lines you need, and mask those.
  3. No identity checks by AI. AI must never decide whether a caller is really the staff member asking for a password reset. Follow your verification process.
  4. No untested commands on live systems. Run AI-suggested scripts on a test machine, with someone who understands every line.
  5. No unapproved tools. Use only the AI tools your organisation has approved for work data. A paid or enterprise plan may have different data terms from a free one; read them, or ask your manager.
  6. No AI agent with keys to everything. If an AI tool can read email or tickets and also take actions, a cleverly written email can try to trick it. This is called prompt injection. Give such tools the smallest access possible.

Kenyan data rules and security context

The Data Protection Act, 2019 covers the personal data in tickets, asset registers and logs. Section 2 defines personal data as information relating to an identified or identifiable person. Section 25 says personal data must be limited to what is necessary, and not transferred outside Kenya unless there are adequate safeguards or the person's consent. Many public AI tools process data outside Kenya, which is one more reason to mask.

Breaches have deadlines. Under section 43, where personal data has been accessed by an unauthorised person and there is a real risk of harm, the data controller must notify the Data Commissioner within 72 hours of becoming aware of it. A data processor must tell the controller within 48 hours where reasonably practicable. If someone pastes a staff list into the wrong tool, report it to your data protection officer at once. For more, read can staff use ChatGPT under Kenya's Data Protection Act?

Attackers use AI too. The National KE-CIRT/CC's cybersecurity report for April to June 2026 (42nd edition) says AI-enhanced phishing became "increasingly convincing", with fake login portals, deepfake videos and voice impersonation. It also notes criminals disguising malware as legitimate AI tools. The CERT urged phishing-resistant authentication, better email filtering and independent checks on sensitive requests before approving them. Your desk should know how to report a national-level incident: the KE-CIRT/CC lists incidents@ke-cirt.go.ke and www.ke-cirt.go.ke.

How FuKazee trains support teams

Our AI training for customer service teams page explains how we plan training around a live support desk. One exercise turns open tickets into a short handover note: what is waiting and what was promised. We use what staff can reach: free AI tools, Microsoft 365 and Copilot, or your help desk. One of our seven modules covers data leakage, hallucinations, prompt injection and policy design. Staff practise on their real kinds of problems, with every personal detail removed. For how to train the whole support function, read how customer service teams should be trained on AI.

Questions people ask

Will AI replace the IT help desk?

AI can answer simple "how do I" questions and draft replies, but someone still has to fix the laptop, check identity and own the outage. It is more useful as a fast assistant to the support officer than as a replacement.

Can I paste an error log into ChatGPT?

Only a short, masked extract, and only into a tool your organisation allows. Remove usernames, emails, server names, IP addresses, paths and anything that looks like a key or token. If the log is long, read it yourself first and paste only the lines around the error.

Is it safe to run a script AI wrote?

Not until you understand every line and have tested it on a machine that does not matter. AI can produce commands that look right but delete, overwrite or open things you did not intend.

Can AI write our IT policies?

It can draft them from your rules, such as an acceptable use policy or a password policy outline. Management approves the final version, and anything legal goes to your lawyer or data protection officer.

Sources

All sources last checked 15 November 2026.

Where FuKazee fits in

FuKazee trains support and service teams to use AI on real desk work, with masked tickets and a clear rule about what never enters a public tool. Training runs live online, or on-site by arrangement, and we plan groups so part of the team stays on the lines. We show the free version of each feature first and favour no brand. We have trained more than 3,000 people since 2022. Microsoft 365 desks can also look at our Microsoft Copilot training and Microsoft 365 training.

To talk about AI training for your IT or service desk team, email lead@fukazee.com

Want guided support instead?

Not sure which programme fits? Tell us who you are and we will point you to the right path: corporate AI training, the ScholarShift Accelerator for postgraduates, or KCSE study plans and tuition.

Rodney MainaFounder & Director, FuKazee

Rodney has run FuKazee's training since 2022 and still teaches inside every programme line: corporate sessions, the postgraduate cohort and the KCSE examiner webinars. First Class Honours in Journalism and Mass Communication, University of Nairobi.