Some problems at work aren’t big enough for a formal complaint, but they won’t go away on their own either. A colleague chases you on three channels for the same thing. Someone on your team hasn’t read the pre-read, again. An agency’s invoices keep coming in over estimate. You know you need to say something, but you also know that if you make it up on the spot, it’ll either come out too soft to change anything or sharp enough to start an argument.
The DESC script is a simple way to plan that conversation. It gives you four sentences to say in order: what happened, what it costs, what you’d like instead and what happens next. Writing them down beforehand means you walk in knowing exactly what you’re asking for.
There’s a catch, though. Two versions of the DESC script are in circulation, and the guides that teach it rarely say which one they’re using. One version’s third step asks you to Specify a change. The other asks you to Suggest alternatives and seek agreement. That sounds like a small difference in wording, but it decides what you say in your third sentence and changes what your fourth sentence is for.
So this post picks one version and explains why. After that come ten filled-in scripts for a planned conversation about a behavior that keeps happening. The other person might be a peer, someone who reports to you, a partner in another team or a vendor. That’s a narrow slice of conflict resolution. You’ve picked the time and counted the incidents, and you have four sentences to get through. Each example also covers the reply you’re most likely to get and what to say back.
After that first reply and your one line back, our assertive communication scripts for when someone pushes back take over.
What is the DESC script?
The DESC script is four sentences, said in order, for raising a behavior that keeps happening with the person who’s doing it. You’ll also see it called the DESC model. Each letter is one step:
- Describe what happened with a count and a date, so it’s a record rather than an opinion.
- Express what it costs you or the work in one sentence.
- Specify one change you’d like and check whether they’ll say yes to it.
- Consequences: say what gets better if it changes and what you’ll do next if it doesn’t.
The order is doing real work. Starting with facts gives the other person something they can recognize before you ask anything of them. The cost tells them why you’re raising it at all. Only then do you ask for something. The last sentence makes it clear what happens either way, so the conversation doesn’t end in a vague “let’s see how it goes.”
The script isn’t new. Yale’s DESC handout for managers names the source: “The DESC script was developed by Sharon and Gordon Bower and is discussed more fully in their book, Asserting Yourself.” So the workplace version is attributed to Sharon and Gordon Bower’s Asserting Yourself (1976).
It also has a second life in healthcare. The US Agency for Healthcare Research and Quality (AHRQ) teaches DESC in TeamSTEPPS, its teamwork curriculum for care teams. AHRQ’s TeamSTEPPS tool page is also precise about where it fits: “The DESC script is a constructive approach to managing and resolving conflict. While used to address both informational and interpersonal conflict, it is most effective with the latter.”
That last line is worth slowing down on. An informational conflict is a disagreement about facts, and it usually ends as soon as someone checks them. An interpersonal conflict is about how people behave. Checking the facts doesn’t settle it, because everyone already knows the facts. That’s what DESC is for: the behavior that keeps happening after the facts are clear.
Specify or Suggest: which version of DESC are you using?
Both versions use the same four letters. If you put them side by side, the first two steps barely differ: both ask you to describe the situation concretely and then say how it affects you. The real differences are in the S and the C. The table shows each step in its source’s own words.
| Step | Bower line, as Yale’s handout words it | TeamSTEPPS line, as AHRQ’s tool page words it |
|---|---|---|
| D | ”Describe the behavior/situation as completely and objectively as possible" | "Describe the specific situation or behavior; provide concrete data.” |
| E | ”Express your feelings or thoughts about the behavior/situation" | "Express how the situation makes you feel/what your concerns are.” |
| S | ”Specify what behavior/outcome you would prefer to happen" | "Suggest alternatives and seek agreement.” |
| C | ”Specify the consequences if this happens. (Both positive and negative)" | "Consequences should be stated in terms of impact on the patient and established team goals; strive for consensus.” |
Look at the S row first. The Bower line makes S a request: you say what you’d prefer to happen. TeamSTEPPS makes S an invitation to work it out together: you suggest options and look for agreement. The C row shifts in the same direction. Bower asks you to spell out the consequences, good and bad. TeamSTEPPS ties them to the patient and the team’s shared goals, and asks you to aim for consensus. That makes sense on a ward, where everyone in the room is working toward the same patient.
(If you go looking for the source, note that AHRQ’s tool page describes its four steps in the text of its DESC graphic rather than in a heading.)
The ten scripts below use the Bower line, with S as Specify. There are three reasons for that.
The first is that one specific request is the easiest thing for someone to say yes or no to. “Suggest alternatives and seek agreement” assumes you share a team and a goal. That holds on a care team, but it’s much shakier with a vendor who answers to a different company.
The second is that every script here is about a behavior that has already repeated. You’re past the point of brainstorming options, so one request keeps the script to four sentences and a single yes or no.
The third is that the useful part of TeamSTEPPS stays in. Every S below ends with a yes-or-no question, so you’re still seeking agreement on the one thing you asked for.
We’ve also made one change of our own, and it’s worth being upfront about it. Both sources treat Consequences as an outcome, something that results from the behavior. We write C as your own next step: something you’ll actually do if nothing changes. That’s our adaptation of the Bower-line consequences, not a rule from either source. We do it because an outcome you announce to someone else can easily sound like a warning, while a step that’s yours alone is just information. The nearest precedent is in the sources’ own samples, which both describe benefits rather than penalties. AHRQ’s sample consequence reads: “A private conversation would be more beneficial to me because I would feel less embarrassed and would be able to ask questions and supply information.” Yale’s sample is a benefit too: “This way we can complete the setup in less time, giving us both a chance to gather our thoughts before we present.”
How to write each letter of a DESC script
The opener before the D
Before you get to the D, plan one sentence that asks for the time and names the topic: “Do you have ten minutes this afternoon? I want to talk about how requests reach me.” Naming the topic up front is fairer than springing it on someone halfway through another conversation, and it lets them arrive ready to talk.
Pick somewhere private. AHRQ’s reason is that a private location “will allow both parties to focus on resolving the conflict rather than on saving face,” and it adds: “Focus on what is right, not who is right.” Put simply, an audience changes what both of you are trying to do. The other person spends their energy not looking bad, and you spend yours winning. Yale’s advice is practical: “Try writing down what you will say and practice it before you talk to the person.” The rest of this post assumes you’re doing exactly that.
D: count it, date it, leave out the motive
AHRQ’s instruction for this step is “provide concrete data.” In practice that means a count and a date. “Three times this month, most recently Tuesday” survives whatever the other person says next, because it’s a record they can check. “You always do this” gets argued with, because “always” is never accurate. Once you’re debating the word “always” you’ve lost the point.
Leave out any guess about why they did it. “You signed it off four minutes after I sent it” is a description. “You didn’t bother reading it” is a guess about motive, and they’ll defend the motive instead of talking about what happened.
E: say what it costs, in one plain sentence
E is usually one sentence about cost or concern at work: what the behavior does to your time, the work or the customer. It tells the other person why this matters, not just that it happened. You don’t have to talk about feelings if that isn’t natural for you at work. But if you’re annoyed, “I’m annoyed” is honest and it works. Stop at one sentence, because a second one tends to turn into a speech.
S: one request small enough to say yes to
Ask for one change, or two at most. Tie it to a time, a place or a channel so you can both tell later whether it happened. “Send it in the queue” can be checked. “Be more organized” can’t, and a request nobody can check usually turns into a second argument about whether it was met.
End the S with a yes-or-no question, such as “Can you do that?” If they say no, don’t carry on to C as if they’d agreed. Adjust the request first, because a consequence attached to a request they’ve just turned down sounds like a punishment.
C: write it as your own next step, not a penalty
Bower’s line says consequences can be “both positive and negative,” and the negative half is where scripts tip into threats. “If this keeps happening, I’ll have to take it further” sounds calm on paper, but it lands as a warning. That means the other person will spend the rest of the conversation reacting to that instead of to your request.
Before you settle on your C, ask yourself four questions:
- Is it something you do yourself, with nobody else’s permission?
- Would you do it anyway, even if you never said it out loud?
- Does it leave their pay, their contract, their reputation and their boss alone?
- Can you say it in a flat voice, the way you’d tell someone the meeting moved rooms?
If the answer to all four is yes, your C is information rather than pressure. Here are two before-and-after pairs we’ve written to show the difference:
- Before: “If the handovers stay this thin, I’ll raise it with your manager.” After: “If they stay thin, I’ll send a written recap after each handover so we’re both working from the same list.”
- Before: “Next time the numbers are late, I’m escalating.” After: “If they’re late on Thursday, I’ll pull them myself and send you what I used.”
The “after” version in both pairs is something you’d sensibly do anyway, and it doesn’t involve anyone’s boss.
10 DESC script examples for the office
Each example below starts with the situation. After that it walks through the four lines as you’d say them, and why each one is worded that way. It finishes with the reply you’re most likely to get and how to answer it. The names are made up, and so are the numbers. When you write your own, use your own counts.
1. A colleague who sends the same request three ways
Nolan is an account executive, and he moves fast. When he needs something from you, he asks on Slack. If nothing comes back in a few minutes, he sends an email. If that sits for a while, he turns up at your desk. It has happened three times this month, each time inside an hour. None of it is rude. But you end up answering the same question in three places, and the two you didn’t answer look as if you ignored him.
Start with what happened, not how you feel about it: “Nolan, three times this month the same request from you has come in on Slack, by email and at my desk, all within an hour. Tuesday’s pricing question was the latest.” Dates and counts matter here because they’re hard to argue with. “You’re always chasing me” invites a debate, but “Three times this month” doesn’t.
Next, say what it costs you in one sentence: “When that happens I lose whatever I was in the middle of, and the places I didn’t reply look like I’m ignoring you.”
Now the request. Keep it to one thing he can say yes to, and give him something in return so it feels like a trade rather than a rule: “Could you send it once, in the request queue? I’ll reply within two hours.”
The consequence is where most people slip into a threat. Here it’s simply what you’ll do next: “If it comes in three places, I’ll answer the one in the queue and tell you that’s where I replied.” You aren’t taking anything away from Nolan. You’re telling him where to look.
He’ll probably push back on speed: “It’s urgent, and you don’t always answer.” That’s fair, and it’s probably why he started chasing in the first place. So rather than argue, give urgent requests their own lane: “Then put URGENT at the top of the queue item. If you’ve heard nothing in two hours, come and find me.” Your request hasn’t changed, but he now has a faster route for the things that really can’t wait.
2. A colleague who signs off your work in minutes
Imogen is a finance analyst, and she signs off your reconciliations before month-end close. The trouble is how fast she does it. Your last two came back with “looks good” within ten minutes of reaching her, and two errors in them turned up at close anyway. You’ve been treating her sign-off as a real second check, and you’re no longer sure it is one.
Open with what you saw, and keep it to things neither of you can dispute: “Imogen, the last two reconciliations came back ‘looks good’ within ten minutes, and two errors in them turned up at close.” This line doesn’t say she didn’t read them. You don’t know that, and if you suggest it she’ll spend the conversation defending herself. The timing and the errors make the point on their own.
After that, explain why it matters to you: “I’m treating your sign-off as a second check, and right now I can’t tell what it covers.” That’s honest without being an accusation, because the problem isn’t that she’s careless. It’s that you’re relying on something you can’t see.
The request should be something small and visible. Put a limit on it so it doesn’t feel like a new permanent process: “For the next two closes, I’d like one line from you saying what you checked. Is that doable?” One line is easy to agree to, and it tells you exactly what her sign-off means.
The consequence has a good side and a fallback, and both are things you’d do anyway: “If that line’s there, I’ll stop re-checking the totals myself. If it isn’t, I’ll send you only the section I want a second look at.” The first half is a real benefit to both of you. The second half isn’t a punishment, because it narrows the job to something she can actually do in the time she’s giving it.
Her most likely reply is about time: “Close week is brutal. I don’t have time to go through all of it.” It’s probably true, but notice that it doesn’t change anything you said. The errors still happened, and you still need to know what she checked. A too-busy reply brings no new fact, so your description and your request both stay. What you can offer is a smaller piece of the same ask: “Then just the intercompany tab. Ten minutes, and the line says you checked it.” It fits inside the time she’s already giving it, and you finally know what that time covers.
3. A colleague who keeps parking your item
“Let’s take it offline.” Gavin is a project lead who has said it about your item in four weekly syncs running, but nothing has happened offline. Each week the vendor-access item gets parked, and each week it slips further down the list while two of your tasks wait on it. You don’t manage Gavin and can’t make him do anything, which is the usual position in conflict resolution for individual contributors. What you can do is ask clearly and control what you do next.
Start with the record: “In the last four syncs, the vendor-access item got moved offline, and there’s been no follow-up after any of them.” Four syncs in a row is hard to wave away, and “no follow-up” is a fact, not a judgment about whether he cares.
The cost is concrete: “It’s blocking two of my tasks, and each week it falls further down the list.”
Your request should be something that can happen in the meeting itself, because “offline” is where the item keeps dying: “Before the meeting ends, can we put it in the notes with an owner and a date?” An owner and a date turn a vague promise into something anyone reading the notes can check.
The consequence has to be something you can do without Gavin’s permission. Adding a line to shared notes fits: “If it gets parked again, I’ll add it to the notes myself, owner TBD, so it stays visible.” You’re not going over his head, just making sure the item doesn’t disappear.
His most likely reply is an easy yes: “Sure, I’ll get to it.” It sounds like agreement, but it’s close to what he’s been saying for four weeks. So ask for the part that’s been missing, which is a day: “Which day this week?”
Gavin says he’ll check his calendar, and that’s where this one stays. Not every DESC conversation ends with a clear commitment, and this one doesn’t.
4. A team member who brings you decisions she can make herself
Joanne reports to you, and last week she brought you six decisions. Every one of them sat inside her own approval limit, so she could have made all six herself. Each one waited on your calendar, and two of them waited a day longer than they needed to.
This script is a little different from the others, because you aren’t asking someone to stop doing something. You’re handing her more room. So the C is there to protect that room. Any disagreement about how she uses it moves into constructive feedback afterward, rather than into undoing her decisions.
Describe it plainly: “Last week you brought me six decisions, and all six were inside your approval limit.” There’s no criticism in that sentence, and there shouldn’t be, because she may well think she was being careful.
The cost is mostly about time: “Each one waited on my calendar, and two of them waited a day longer than they needed to.”
The request gives her the authority and a simple way to keep you informed: “Decide anything inside your limit and list it in Friday’s summary. Are you fine with that?” The Friday summary matters because you’ll still see what she decided, so she isn’t left on her own with it.
The consequence is a promise, and it’s the most important line in this script: “I won’t reverse a call you make inside your limit. If I’d have gone another way, I’ll tell you afterward and before the next one.” It’s hard to decide anything when you suspect the decision might be overturned, and this line tells her it won’t be.
Add your offer alongside it: anything above her limit or anything she’s unsure of still comes to you, with her recommendation attached. That keeps a route open for the harder calls and gets her used to having a view on them.
Her most likely reply is a worry rather than pushback: “What if I get one wrong?” Answer it by pointing to the safety net you’ve already built: “Then we’ll see it on Friday, and we’ll adjust the limit or the guidance. That’s what the summary is for.” A wrong call becomes something you fix together, not something she carries alone.
5. A team member who hasn’t read the pre-read
Callum hasn’t opened the pre-read for the last three weekly reviews. About fifteen minutes of each meeting has gone on catching him up, so the decisions the review is actually for get squeezed into the end.
Describe it with the numbers: “For the last three reviews, the pre-read hadn’t been opened, and we spent around fifteen minutes catching up at the start.” Saying “the pre-read hadn’t been opened” rather than “you didn’t read it” keeps the focus on what happened, not on him.
The cost falls on more people than just you: “That comes out of everyone’s time, and the decisions get squeezed into the last ten minutes.”
The request gives him two acceptable options, which makes it easier to say yes: “Read it beforehand, or tell me the evening before if you can’t, so I can move your slot. Can you manage that?” You aren’t asking for perfection, just asking him not to let it come as a surprise.
The consequence is only a change to the running order: “If it hasn’t been read and I haven’t heard, I’ll start with the others and take your items last.” Say it the way you’d announce an agenda order, because that’s all it is. Nobody is being punished, and the meeting just doesn’t wait for anyone.
His most likely reply is about workload: “My week was all client work.” That’s a new fact, and it’s worth taking seriously because it hints that the pre-read itself might be part of the problem. So change the request once: “Is it the length? I can send a one-page version on Monday and keep the full deck as backup.”
6. A team member whose client summaries don’t match the source
Three times this month, figures in Leah’s client summaries didn’t match the source sheet they came from. The latest was the renewal numbers in last Tuesday’s summary. These summaries go to clients, so a mismatch is more than a tidiness problem.
Describe it with the count and the most recent case: “Three summaries this month had figures that didn’t match the source sheet. The renewal numbers in last Tuesday’s were the latest.” Naming one specific example gives her something to go and look at, rather than a general complaint about accuracy.
The cost explains why this matters more than it might seem: “The client reads the summary, not the sheet, so a wrong number becomes the number.” That framing makes it clear you’re worried about the client, not about catching her out.
The request is small and easy to check: “Please link the sheet tab and its date in each summary. Can you do that?” A link and a date on every summary let anyone trace a figure back to where it came from.
The consequence is a short check with a clear end: “For the next month, I’ll spot-check Tuesday’s summaries before they go. Two clean weeks and I stop.” The exit is the important part. A spot-check without one feels like being watched with no end in sight, but an exit tells her exactly what it takes for the checking to stop.
Her most likely reply explains the gap: “The sheet changed after I pulled it.” That’s a new fact, and it might be the whole problem. If the numbers moved after she copied them, the fix isn’t for her to be more careful but to agree on when the numbers are frozen. So change the request: “Then let’s set a cut-off. You pull at 10am Tuesday and the summary says ‘as of 10am.’ Late changes go in the next one.”
7. A sales lead who promises custom work before product hears about it
You’re in product. Kieran is a sales lead, and on three calls this quarter he’s promised a prospect custom work before anyone in product knew about it. Each time, your team has had to choose between building something that wasn’t planned and going back to the customer to undo the promise. Partners in another team don’t share your roadmap or your targets, and that’s most of what makes cross-functional collaboration hard. Kieran isn’t trying to cause trouble. He’s trying to close a deal.
Describe what happened without commenting on his judgment: “On three calls this quarter, a custom feature was promised to a prospect before it reached product.”
Put the cost in terms he cares about, which is the customer: “We then either build something off the roadmap or walk the promise back, and both cost us with the customer.”
The request asks him for a quick check and gives him a fast answer in return: “Before you commit to anything off the roadmap, message the deal channel and I’ll answer the same day. Will you do that?” The same-day promise matters, because a sales lead won’t wait a week for an answer while a deal goes cold.
The consequence is something useful you’ll do for him regardless: “Before your next call, I’ll post what’s on the roadmap in the deal channel, so you can see it before you’re asked.” There’s no penalty in it at all, and it just means he walks into calls knowing what product can deliver.
His most likely reply is about timing: “If I wait, I lose the deal.” That’s a fair worry, and it tells you something new about how his calls go. So adjust the request to fit. He doesn’t have to make the prospect wait, but he does have to avoid committing on the spot: “Then don’t wait on the call. Say ‘Let me confirm and come back to you,’ and I’ll get you an answer the same day.”
8. A support lead whose team escalates by direct message
“If I ticket it, nobody looks at it till tomorrow.” Megan runs support while you’re in engineering, and she’s partly right. Last week her team sent six escalations straight to engineers by DM, and two of them were never ticketed. You can see why her team does it, but you can also see what it costs.
Describe it with the numbers: “Last week, six escalations came to engineers by DM, and two of them never got a ticket.”
The cost lands mostly on her team and their customers: “Those two have no record, so on-call can’t pick them up and nobody can tell you whether they’re fixed.” Framing it this way shows that the DMs are working against what she wants, which is problems getting fixed quickly.
The request keeps the DM but changes what it’s for: “I’d like your team to queue it first and use the DM only to send the ticket link. Is that workable on your side?” Her team still gets the speed of a direct message, and you get a record.
The consequence is a rule for your own team, which is entirely in your hands: “From Monday, an engineer will ask for the ticket before starting on a DM. I’ve told my team that’s the rule.” You aren’t telling her team what to do, only what yours will do.
Because her point about speed is a fair one, make an offer alongside it: P1 tickets get a response within an hour. That answers her original complaint directly.
Her most likely reply is about the form itself: “The ticket form takes five minutes, and the customer is on the line.” That’s another fair point, and it’s a new fact you can act on. So rather than repeating the rule, offer to fix whatever is slowing her down: “Paste the template here and show me which field slows you down. If it’s one we can drop, we’ll drop it.”
9. An agency that bills over estimate without warning
Ewan is an account director at your agency, and twice now he’s billed hours over the agreed estimate without telling you first. Both times, you found out when the invoice arrived. This conversation happens by email rather than face to face, so the four lines become four short paragraphs. The rest of the advice on disagreeing in writing applies too.
Here’s how the email reads:
Hi Ewan,
The last two invoices each billed hours over the agreed estimate, and in neither case did we hear about it before the invoice arrived.
That leaves me explaining a budget overrun I could have planned for.
When a piece of work reaches 80% of its estimated hours, please flag it and ask before going over. Does that work for your team?
Until then, I’ll go through any over-estimate invoice line by line with you before I approve it.
Each paragraph does one job. The first is the record: two invoices, no warning. It doesn’t accuse the agency of padding hours, because you don’t know that. What you can show is that nobody warned you. The second is the cost, and it’s about your position, because you’re the one who has to explain the overrun to someone else.
The third paragraph is the request, and the 80% mark is what makes it work. It’s a clear point that has either been reached or hasn’t, and it gives Ewan time to raise the issue while there’s still room to decide what to do. The last paragraph is your next step. Going through an invoice line by line before approving it is something any client can reasonably do. It doesn’t hold back payment or mention other agencies. It just means overruns get looked at properly.
Ewan’s most likely reply pushes some of it back on you: “Scope grew because your team kept adding changes.” Some of that may be true, so don’t deny it. Split the hours instead: “Some of it did. Let’s split the hours: anything on our change-request list is scope growth and I’ll treat it that way. The rest is overrun, and that’s what the 80% flag is for.” That separates the hours you asked for from the hours nobody warned you about.
This is a negotiation now, and it isn’t settled. Ewan says he’ll look at the list, but the next invoice hasn’t arrived yet.
10. A recruitment agency whose profiles miss the brief
“We’re sending you the closest we can find.” Fiona is your account manager at a recruitment agency, and she has a real point. But four of the last six profiles she sent were missing two of the must-haves in the brief, and every one of them took up your time.
Describe it with the count: “Four of the last six profiles were missing two of the must-haves in the brief.” Keep it about the profiles rather than about how hard she’s trying.
The cost includes the person you answer to: “Each one costs me a screen, and the hiring manager is starting to doubt the shortlist.”
The request asks for a check she can do before she sends anything: “Please check each profile against the three must-haves before you send it. Would that be possible?”
The consequence is simple and specific: “If a profile misses one, I’ll send it back with the missing must-have named.” That helps her too, because naming what’s missing tells her exactly what to look for next time.
Her most likely reply is the honest one: “The market’s thin. We’re stretching on some of these.” That’s a new fact, and it should change the request. If the right people aren’t out there, insisting on a perfect match may just mean fewer profiles. So ask her to be open about the gaps instead: “That’s useful to know. Keep sending them, and mark which ones are stretches and which must-have they miss.” You still see every profile, but now you know which ones are stretches before you spend a screen on them.
If you want to hear how a line like this sounds before you send it, you can rehearse it with Merlin. Merlin can play Fiona, reply included.
What to say when the reply comes
Most first replies fall into four kinds: “I didn’t know,” “that’s not what happened,” “I’m too busy,” and “it’s urgent.” You’ve seen most of them in the examples above. How you answer comes down to one question: does the reply tell you something you didn’t know?
If it brings a new fact, like a sheet that changed after it was pulled or a thin market, change your S once. Your request was based on what you knew when you wrote it, and now you know more. Callum and Leah both brought new facts like that. So did Kieran, Megan and Fiona.
If the reply is a deflection, don’t change anything and restate the S in the same words and offer the next step. A too-busy reply counts as a deflection, because being busy doesn’t change what happened. That’s what happened with Imogen: the ask stayed the same and you offered a smaller piece of it. What you shouldn’t do is add a new reason. Every extra reason is something new to argue with, and the conversation drifts away from the one change you asked for.
The script only plans the opening. Once you’ve had the first reply and given your one line back, the scripts for holding your ask under pushback pick up. They have you acknowledge what was said, repeat the ask and offer the next step. If the reply turns hot, stop the script and use de-escalation techniques instead, because a planned script doesn’t survive a raised voice.
When the DESC script backfires
One incident is not a pattern
The D step needs a behavior that has repeated. If you use DESC on a single slip, you’ve built a case file out of a mistake. The other person hears a count of one delivered with the structure of a count of five, and the formality itself feels like an overreaction.
A first miss just needs a plain question: “The summary had the old renewal figure. Can you fix it before it goes out?” Keep DESC for the third time.
A power gap changes what S means
When a manager uses DESC with someone who reports to them, Specify stops sounding like a request and starts sounding like an instruction. Joanne and Callum can’t really say no to their manager’s yes-or-no question, so you may get compliance without agreement. They say yes because no doesn’t feel like an option. You only find out which one you got weeks later, when the behavior either sticks or quietly comes back.
It’s awkward the other way round too. When you use DESC with your own manager, a dated and counted D can look like a case being built against the person who writes your review. Raising something upward works better when you tie your ask to a trade-off they care about. Our post on pushing back on your manager covers that, and this one doesn’t.
When the C step escalates
A C that reaches their boss, their contract, their pay or anything else outside your control reads as a threat. That stays true however flatly you say it.
Written the obvious way, three of the Cs above would have done exactly that. The obvious C in example 7 would be “I’ll talk to your customer directly before the next call.” Example 9’s would be “We’ll hold payment on the overage until it’s explained.” Example 10’s would be “If this continues, we’ll look at other agencies.” Each one sounds reasonable on the page, but each one is pressure, and the other person will hear it that way.
There’s a reason those three are the ones that go wrong, and the peer scripts aren’t. All three are with someone outside your team. TeamSTEPPS frames C in terms of shared team goals, which works when everyone is on the same care team. A vendor or another department doesn’t share a team with you, so a consequence written around goals slides into pressure.
If the behavior carries on after one round, do the thing you said you’d do. Do it once, and don’t run the script again. Your C already covered what happens next, and repeating the whole script just turns it into nagging.
One note on scope. AHRQ says DESC is “especially appropriate for major conflicts such as hostile or harassing behaviors” on a care team. But an office is different. Harassment, discrimination or a safety concern there goes to HR first, not into a four-sentence script.
DESC also has nothing to say about the person who agrees and then doesn’t change. Gavin said yes, and Ewan said he’d look. The only part of the script that still works after a yes like that is the C, because it’s the only part that doesn’t need them.
Run it past Merlin before you say it
Yale’s advice holds for every script here: write it down and say it out loud before the real conversation. Saying it out loud is where you’ll hear the C that sounds like a threat, or the D that’s really a guess about motive. Both can look fine on paper and sound quite different when spoken.
Merlin runs natively in Slack and Microsoft Teams, so you can practice without leaving either. Give Merlin your four lines and it can play the other person, likely reply included. You can then try your next line against it until it holds. When you start the practice yourself, it stays private to you.
Try your DESC script with Merlin before the conversation it’s for.
Pick the one repeated behavior you’ve been putting off raising, and write the S line first. It’s the hardest of the four to get right. But once you know exactly what you’re asking for, the other three lines are much easier to write.