How to Run a Heist Campaign in D&D (Without Scripting Every Failure)

Every heist campaign starts the same way: a target, a plan, and a party convinced the plan is airtight. It never is. Somewhere between "we disable the wards" and "we're inside," a guard changes a patrol route, a dice roll goes bad, or a player improvises something nobody prepped for, and that's the moment a heist campaign either comes alive or falls apart.

Most guides to running a heist focus on the planning phase: crew roles, floor plans, the big score. That part is the easy half. The part that actually determines whether the table remembers the session is what happens when the plan stops mattering, and that's the half most GMs walk in without a real answer for.

How to Run a Heist Campaign in D&D (The Short Version)

Build the Job Backward From the Score

Start with what's being stolen, not with the building it's locked inside. The target decides everything else: how it's guarded, who wants it kept safe, and what happens in the city after it's gone.

A few questions worth answering before you draw anything:

• Why is this specific thing guarded this way? A sentimental family heirloom gets a single loyal guard and a locked drawer. A relic that can level a district gets wards, a standing patrol, and a owner who checks on it personally. The security should match the stakes, or players will clock the mismatch immediately.
• Who notices it's missing, and how fast? This sets your clock. A merchant who checks his safe nightly gives the party hours. A crime lord who only opens the vault once a season gives them weeks, which changes the whole shape of the job.
• What does the target's owner do about it? This is the part that turns a one-shot heist into the start of a campaign. A stolen relic with a furious, resourceful owner chasing it down is a built-in antagonist for the next several sessions.

Give the Planning Phase Real Weight

The planning session is where most of the heist's actual decisions get made, and it deserves its own time at the table instead of a quick rundown before the "real" session starts.

Recon should answer specific questions, not just reveal a map. Instead of handing over a floor plan, let scouting rolls answer things like "how many guards work the night shift" or "does the vault door have a second lock." Specific answers to specific questions make players feel like their planning actually shaped what they know, instead of being handed the solution.

Let the party gather tools and favors, not just information. A forged invitation, a borrowed uniform, a bribed guard, a rope that can take real weight. These give players concrete things to spend during execution, and spending a resource they worked for feels different than rolling a check cold.

Expect the plan to be wrong, and don't correct it. Players will build a plan around partial information, because that's what recon gives them. If a player asks "is this definitely going to work," the honest answer is "you don't know yet," not a reassurance that locks in the GM's hidden answer.

Heist Crew Roles That Actually Matter at the Table

Forget matching character classes to heist archetypes one-to-one. What a heist crew actually needs is four functions covered, and most parties can cover all four with three or four characters.

Someone who gets in quietly. Lockpicking, stealth, finding the unguarded entrance. This is the role everyone assumes a heist needs, and it's necessary but rarely sufficient on its own.

Someone who talks their way past people. Guards, clerks, nosy neighbors. A face character who can bluff, persuade, or disguise their way through a human obstacle saves the crew from having to solve every problem with a lockpick or a sword.

Someone who handles obstacles that aren't people. Wards, wired alarms, a vault mechanism that needs more than brute force. This role matters most in jobs where the security isn't a guard, it's a system, and systems don't respond to intimidation.

Someone who covers the exit. A heist that goes loud needs a way out that isn't "we fight through everyone between here and the door." A getaway plan, a diversion, a bribed guard at the back gate. This is the role parties forget to assign until they desperately need it.

The Twist: Plan for the Complication, Not the Perfect Heist

Here's the part most heist advice skips past. A heist campaign isn't memorable because the plan worked. It's memorable because the plan almost worked, something broke at the worst possible moment, and the crew improvised their way through it anyway.

That means your job as GM isn't to prepare the "correct" outcome of the job. It's to prepare two or three ways the job can break, without deciding in advance which one happens or how the party handles it.

Useful complications to have ready, loosely:

• A detail recon missed. The guard shift changed. There's a second lock nobody scouted. This rewards good planning without punishing it, since the party still knew most of what they needed.
• A roll goes badly at the worst moment. A failed stealth check mid-job, a botched lie to a guard who's now suspicious. This is where a tense scene actually becomes tense, instead of the party sailing through on a plan that already accounted for everything.
• An NPC does something the party didn't predict. The guard captain decides to do an unscheduled walkthrough. The target shows up early. NPCs acting on their own motivations, not the party's script, is what makes a heist feel like a living situation instead of a puzzle with one solution.

What kills this is a GM who treats deviation from the plan as a mistake the party made, rather than the actual game. If the party's improvised fix gets punished as hard as, or harder than, simply failing, players stop improvising and start grinding for the "safe" answer, which is the exact dynamic that makes heists boring.

Where an AI Game Master Fits Into a Heist Campaign

This is the part of running a heist that's genuinely hard for a human GM to do well on demand: react fast and fairly to whatever the party actually tries, every single session, without falling back on the plan you prepped because improvising live is tiring.

A heist campaign asks more of real-time improvisation than almost any other format. The planning session can be prepped in advance. The execution session can't, because the entire point is that players are reacting to a situation that's reacting back to them. That's exactly the kind of moment-to-moment responsiveness an AI dungeon master is built for.

StoryRoll's D&D mode keeps the full ruleset, so a heist still plays like a real 5e session: skill checks for the lockpicking, persuasion rolls for the talk-your-way-past-the-guard moment, initiative if the job goes loud. The AI Game Master tracks what the party has actually learned during recon and reacts to the plan they built, rather than steering them back toward a version of the job it had in mind. When a player tries something that wasn't in anyone's plan, including the GM's, the response comes from what's actually happening at the table instead of from a script that needs a detour.

It also solves a scheduling problem specific to this format: heists work best run close together, planning session into execution session, while the plan is still fresh in everyone's head. A human GM's calendar doesn't always allow that. A crew that wants to run the job the same week they finished recon can just do it.

Heist Campaign Mistakes to Avoid

Scripting the "correct" route and punishing everything else. If there's one path through the vault that works and every other approach gets harder rolls or worse outcomes, players will feel railroaded the moment they notice.

Giving recon all the answers up front. If the planning session hands over a complete floor plan and guard schedule with no gaps, there's nothing left for the party to miss, which means there's nothing left to go wrong in a way that feels earned.

Letting nothing go wrong, ever. A heist where the plan executes flawlessly isn't a heist, it's a cutscene. The tension comes from things breaking and the crew handling it.

Skipping the planning session entirely. Jumping straight to "you're at the vault" cuts out the part where players build investment in a plan they made themselves, which is half of why the eventual complication lands.

Making every NPC in the building a non-factor until they're relevant. Guards, clerks, and the target's owner should feel like they have routines and motivations before the heist starts, not like set dressing that only activates when the plot needs tension.

Frequently Asked Questions

What is a heist campaign in D&D?

A heist campaign is a D&D format built around planning and executing a theft, usually from a well-guarded target like a vault, a noble's estate, or a crime lord's vault. Unlike a dungeon crawl, the core loop is scouting, planning, and improvising around a specific objective rather than exploring a space room by room.

How do you structure a D&D heist?

Work backward from the score. Decide what the party is stealing and why it's guarded the way it is, then build the planning phase (recon, gathering intel, acquiring tools) and the execution phase (the actual break-in) around that target. Most heists run as one planning session and one or two execution sessions, not a full campaign arc.

What roles should a heist party have?

A workable heist crew covers four functions: someone who gets in quietly (a rogue or scout), someone who talks their way past people (a face or social character), someone who handles obstacles like locks, wards, or wiring (a specialist), and someone who can get the crew out if it goes loud (muscle or a getaway plan). A party doesn't need one character per role; most characters can cover two.

How much should the GM plan in advance for a heist?

Plan the target, not the outcome. Know the building's layout, the guards' patrol routine, and two or three ways the plan can break, but don't script what players actually do. Heists fail as campaigns when the GM prepares one correct path through the vault and treats every other player choice as a wrong answer.

Can an AI Game Master run a heist campaign?

Yes, and it's a good fit for the format's biggest demand: heists need a GM who can improvise fast when the plan breaks, which happens almost every session. StoryRoll's AI Game Master reacts to whatever the party actually tries, in real time, instead of steering them back toward a script, and it can run the planning session and the execution session back to back without the usual gap between prep and payoff.

What's the biggest mistake when running a D&D heist?

Treating the plan as the thing that has to survive, instead of the complication. A heist campaign isn't fun because the plan works. It's fun because the plan almost works, something breaks at the worst moment, and the crew improvises their way out. GMs who punish deviation from the plan, or who never let anything go wrong, both drain the tension that makes heists worth running.

Try These Free Tools

• NPC Name Generator: name the guards, clerks, and fences a heist crew runs into without stopping mid-plan.
• Loot Generator: fill the vault with something worth the risk before the party ever gets inside.
• Skill Check Reference: set fair DCs for the lockpicking, persuasion, and stealth checks a job actually runs on.