Home › Topics › Discovering Shadow IT Through Network Data › The Diplomatic Conversation

The Diplomatic Conversation: Talking to Shadow IT Users


The report is done. Twenty-three machines in the design department have been syncing to a file-storage service nobody approved, for over a month, at a few gigabytes a day. You know which team. You know roughly which desks. What you do in the next hour decides how the design team behaves next time. They may become the group that tells you about the next tool before adopting it. Or they may learn to hide things from IT more carefully. The technical part of shadow IT discovery is over. The human part has just started, and it decides whether any of the technical work was worth doing.


This guide is about that conversation. It is written for the admin who is good with switches and uncomfortable with confrontation, which describes most of us. The good news is that the right approach is not confrontation at all. It is a short, honest, curious conversation with a colleague who solved a problem, followed by an agreement about what happens next. I have had this conversation badly and I have had it well, and the difference was never the facts I brought. You will learn how to prepare, how to open, and what to listen for. You will learn how to explain risk without lecturing, what to offer, and when to escalate. You will also learn how to write down the outcome so it counts.


You will also learn how to avoid the one thing that poisons these conversations: the surveillance vibe. You noticed a service. You did not read anyone's files. Keeping that distinction clear, in your own head and in what you say, is most of the job.


In this guide

  • The mindset: they solved a problem you did not
  • Before you walk over
  • Opening the conversation
  • Listening for the unmet need
  • Explaining the risk, briefly
  • Offering the sanctioned path, or admitting there is none
  • Agreeing on next steps and writing them down
  • Avoiding the surveillance vibe
  • When to escalate
  • The scenario table
  • Where to go next

The mindset: they solved a problem you did not


Start from the most likely truth. The person in front of you had a job to do, the official route was slow, missing, or unknown to them, and they found a way. From their side, they were resourceful. Nobody sets up a project board to defy IT; they set it up because the project needed a board and there was not one. If you walk over believing you are correcting a rule-breaker, everything you say will carry that belief. The person will hear it no matter how carefully you phrase things.


So the operating rule is curiosity before enforcement. Curiosity is not a tactic to soften people up; it is the correct stance. You do not yet know what the tool is for or how important it has become. You do not know whether the right outcome is to remove it, replace it, or adopt it. A finding tells you a service exists. Only the conversation tells you what it means. Enforcement, when it is needed at all, comes after you understand the situation, not instead of understanding it.


There is a practical payoff too. People who are treated as colleagues in this conversation tell you about the next tool before they adopt it. People who are treated as suspects do not, and your census gets less honest every time. Every conversation is also a small advertisement for how IT behaves, and the audience is larger than the one person at the desk.


Before you walk over


Five minutes of preparation turns an awkward visit into a productive one. (Zero minutes of preparation turns it into a second visit.) Run through this list:


  1. Know exactly what you found, and only that. The service or device, how many hosts, how long it has been in use, roughly how much data moves. Not a list of anyone's queries, not the names of files. You should be able to say what you know in one sentence.
  2. Know the sanctioned alternative, or know that there is none. If there is an approved tool for the same job, know how to get someone started on it today. If there is not, be ready to say so plainly.
  3. Decide what you are asking for. Information? A migration by a certain date? Nothing yet, pending a decision? Walking over without an answer to this means the conversation ends with vague mutual nodding.
  4. Check the data sensitivity. If the team handles regulated or contractually protected data, the conversation may not be yours alone to have. In that case, see the escalation section before you go.
  5. Pick the right person and time. Choose the team lead or the person who set the tool up, at a quiet moment. Speak in person or by call rather than by a message that lands like a summons.

Opening the conversation


The opening should do three things in the first thirty seconds: say what you noticed, say how you noticed it, and ask a genuine question. Something like this:


"Hi — I've been going through the list of cloud services our network talks to. One that came up is a file-sync service that a bunch of machines in your team are using. It's not one we've set up officially, so I wanted to understand what you're using it for and see what we can do. What were you trying to solve?"


Notice what that does. It names the service, not a person. It explains that the source is a network-level census of services, which is both true and reassuring. It says "not set up officially" rather than "unauthorized," which is a word people hear in capital letters. And it ends with the single most useful question in this whole guide: what were you trying to do? That question is impossible to hear as an accusation, and its answer tells you almost everything you need to know.


For hardware, the same shape works at the desk: "I'm tracking down a few devices the network sees that aren't in our inventory. One of them is on the jack under your desk — is that a storage box? What's it doing for you?" You are asking about the device and its purpose, not about the person's judgment in buying it.


Listening for the unmet need


Then stop talking. The answer to "what were you trying to do?" is where the value is, and it is usually one of a small number of things:


  • "We needed to share big files with a client and email bounces them." A missing capability. You may not have a sanctioned answer, and you should find out whether you do before promising one.
  • "The shared drive is unusable from home." A performance or access problem with the sanctioned tool. The shadow tool is a symptom; the drive is the problem.
  • "We asked for more storage months ago and never heard back." A procurement or ticketing failure. The finding is partly a complaint about IT, and a fair one.
  • "It's what I used at my last job." Habit. Often the easiest case, because the person has no particular attachment to the tool, only to the workflow.
  • "I didn't know that counted as something IT needed to know about." A visibility gap, common and completely sincere. Also the easiest to fix for next time.

Whichever you hear, repeat it back in your own words so the person knows you got it. For example: "So the real issue is getting large files to outside people quickly." Now you are two colleagues looking at the same problem, which is the posture you want for the rest of the conversation. Resist the urge to jump to the risk speech the moment you have identified the need. The speech will keep. Ask one more question first: how much does the team depend on this now? A tool that holds a month of work needs a migration plan. A tool someone tried twice needs nothing but a better default next time.


Explaining the risk, briefly


Now you earn the right to explain why it matters, and the rule is one concrete risk, in two sentences, aimed at their situation. Not a list. Not a policy citation. Pick the consequence that would actually hurt this team and say it plainly:


"The thing that worries me is that nothing on that service gets backed up by us. If the account gets closed or someone leaves with the login, the team's files are just gone, and we'd have no way to get them back."


Or, for a personal router: "The problem with that router is that anything connected to its Wi-Fi is inside our staff network without any of the protections the rest of the building has. It's basically a side door we didn't know about."


Two sentences, then stop. People understand risk perfectly well when it is specific to them, and tune out when it is general and long. If they push back — "but it's been fine for a year" — agree that it has. Then note that the risk is exactly the kind that is fine until the day it is not, like a backup nobody has tested. Do not escalate the rhetoric. You are not trying to win; you are trying to be understood.


Tip: never lead with "it's against policy." Even when it is, policy is a reason to have the conversation, not a reason that persuades anyone. Lead with the concrete consequence and mention policy only if someone asks what the rules are.


Offering the sanctioned path, or admitting there is none


This is where the conversation succeeds or fails, and the honest version beats the confident one. Three possibilities, each with a right response.


There is a sanctioned tool that does the job. Offer it, and offer to help: "We actually have a way to send big files to outside people. I can set your team up this afternoon and show you how it works." If the sanctioned tool is worse in some way, say so and say what you will do about it. People will tolerate a slightly clumsier tool if you are honest about the trade and it fixes the backup problem.


There is no sanctioned tool, but there could be. Say that plainly: "We don't have anything for that right now. What you found might actually be the right answer. Let me get it looked at properly so we can either approve it or find something equivalent." This is the path that turns a finding into an improvement, and it is more common than admins expect. The sibling guide on building a shadow IT program covers the fast evaluation lane that makes this promise keepable.


There is no sanctioned tool and there will not be one — because the data cannot leave, because of a contract, because of the budget. Say that too, and say why: "That data can't go to a consumer service at all because of our agreement with the client. I know that's a pain. Here's what we can do instead, and here's what I'll push for." An honest "no, and here is why" is respected. A vague "we'll look into it" that never resolves is how the same finding reappears in three months.


At Harbourview Clinic the scheduling team ran its whole staff rota on a free calendar-sharing service. The first person IT sent opened with the acceptable-use policy. The team lead asked which clause covered "the shared drive does not work on the ward tablets," and the meeting ended there. The second visit opened with "I need to understand what this does for you." It ended forty minutes later with a list of six things the rota needed. The sanctioned platform already did five of them, and there was a ticket for the sixth. The rota moved within a month. The policy document, it turned out, had been read by four people, three of them in IT.


Agreeing on next steps and writing them down


End with something specific that both of you can repeat back: who does what, by when. "We'll migrate to the approved tool by the end of the month and I'll help move the files" is an agreement. "We should probably sort this out" is not. Then record it, because a conversation that is not written down did not happen as far as the next review is concerned. A short note is enough:


Finding:      file-sync service, design team, 23 hosts, in use ~6 weeks
Spoke with:   team lead, in person
Need:         shared working folder reachable from home; office drive too slow over VPN
Risk stated:  no backup; data lost if account closes or owner leaves
Sanctioned:   yes - company file platform, external-share feature
Agreed:       IT sets up team space this week; team migrates by month end;
              IT to investigate VPN performance separately (ticket opened)
Status:       tolerated until migration date, then review
Follow-up:    check service host count after migration date

The note records the service, the team, the need, the agreement, and the follow-up. It does not record anyone's browsing, any file names, or any judgment of the person. This is the record that feeds the program's metrics and the next review. It is the evidence, if it is ever needed, that IT behaved reasonably.


Avoiding the surveillance vibe


Everything above depends on the person believing that you noticed a service, not that you were watching them. That belief has to be true, and it has to be visible. A few habits keep it that way:


  • Talk about services and devices, never about content. "A file-sync service" is fine. "The folder called Q3 proposals" is not, even if you could have seen the name somewhere. If you know it, do not say it.
  • Explain the source once, plainly. "We look at which cloud services the whole network talks to, by service and how many machines — not what anyone does on them." Most people are relieved to hear this, and it is the truth.
  • Never bring a printout of queries. Bring one sentence. The moment a list of someone's activity appears on a desk, the conversation is about surveillance no matter what it is called.
  • Do not go to the manager first. Going over someone's head before speaking to them turns a colleague into a defendant. Managers come in at escalation, and only when needed.
  • Say what you will not do. "I'm not here to take anything away today" costs nothing and changes the whole tone.

If you find yourself wanting to mention something you saw in order to make a point land harder, that is the moment to stop. The point will land fine on its own, and the relationship will survive.


When to escalate


Most conversations end in agreement and need no one else. A few do not, and it helps to know in advance which situations are not yours to settle at a desk:


Regulated or contractually protected data. Suppose the service holds medical, financial, personal, or client-protected data. In that case, the question of what happens next belongs to whoever owns compliance or the client relationship. The situation may carry notification obligations you cannot judge. Have the conversation to understand the situation, then hand the decision up promptly and say that you are doing so. Do not promise an outcome you cannot deliver.


Repeat behavior after an agreement. The migration date passed, the host count did not fall, and a friendly reminder changed nothing. At that point the finding is no longer about an unmet need. It is about an agreement that was not kept, and that is a management matter. Escalate with the record: the finding, the conversation, the agreement, the follow-up. Facts, not frustration.


Refusal. Someone declines to discuss the tool, declines to move, or tells you it is none of IT's business. Stay calm, say that you will need to raise it, and do. Refusal is rare, and it usually means the person has been burned before. It still has to go up.


Escalation is not a failure of diplomacy. It is what the diplomacy was protecting: your ability to say, honestly, that you tried the reasonable thing first. Keep the escalation itself factual and short, and keep it about the data path, not the person.


The scenario table


Four findings come up more than any others. This table pairs each with the opening question, the risk sentence, and the approach that tends to work.


FindingOpen withThe one risk to nameRecommended approach
Personal cloud storage syncing work files"What's the folder for — working from home?"No backup; gone if the account closes or the person leavesOffer the sanctioned storage with the same convenience; help migrate; agree a date; check host count afterward
Team project tool nobody approved"What does it do that our tools don't?"Plans and customer names outside retention; no offboarding of accountsOften a candidate for adoption. Route to fast evaluation; tolerate pending decision; do not block mid-project
Personal router or hotspot bridged in"Is the Wi-Fi bad in here? What's the router for?"Side door into the staff network past every controlHighest urgency of the four. Fix the underlying Wi-Fi need quickly; the router comes out, with the reason explained, not just the rule
Mystery NAS under a desk"Is that a storage box? What's on it?"Unpatched device; team's data on one disk with no backupFind out what depends on it; provide managed space; migrate; then retire the box, with the owner present

One walkthrough for the row that goes wrong most often. You find the project tool and tell the team it has to go by Friday. You discover on Thursday that it holds every task for a product launch next week. Now you are the person who broke the launch. The better sequence is the one in the table: understand first, then tolerate while a proper evaluation happens. If it must be replaced, plan the replacement around the team's calendar rather than yours. Blocking is occasionally necessary. Blocking as an opening move almost never is.


Where to go next


You now have a conversation that starts with curiosity, names one concrete risk, and offers an honest path. It ends in an agreement and leaves a record. You know when it stops being yours to settle. The findings that start these conversations come from finding unsanctioned SaaS and finding unofficial hardware. Making those promises keepable — the fast evaluation, the published catalog, the follow-up review — is the subject of the program guide. All of it lives under Discovering Shadow IT Through Network Data. The report is still on your screen and the design team is still at their desks. Go and ask what they were trying to do.


Put this into practice with SysPeep Monitor

The conversation in this guide is human work no software does for you. But the facts you bring to it come from the network. The follow-up is easier when the device side is already visible. SysPeep Monitor auto-discovers the devices and interfaces on your network and polls them over SNMP. A storage box that was supposed to be retired by month end is easy to check on. The free trial installs on any Windows machine in minutes.

Download Free TrialLearn About SysPeep