Duplicate work: how teams end up solving the same problem twice
Why teams end up doing the same work twice, how AI chats and scattered tools make it worse, and six simple habits that stop duplicate work early.
Duplicate work happens when two people solve the same problem without knowing the other has already started or finished. It is rarely laziness: it comes from work that is hard to see, decisions that live in private places, and tools that make it cheap to start from scratch.
It is also common. Atlassian reports that 50% of knowledge workers have unknowingly worked on duplicate projects, and that a quarter of the workweek goes to searching for information. This post covers how it happens, how to spot it, and what to change.
What duplicate work looks like
Most duplicate work is smaller and quieter than two teams building the same product:
Two product managers each build a churn analysis in the same quarter, with different definitions of "churned."
An engineer writes a CSV import helper that already exists in another service.
Sales and marketing each write a competitor comparison, and the two disagree on pricing.
A new hire spends two days rebuilding an onboarding checklist that ops wrote last year.
Three people on the same team each refine their own prompt for writing release notes.
Each case wastes hours. The bigger cost is the conflict that follows: two versions of the truth, and a meeting to decide which one wins.
Why teams end up solving the same problem twice
Nobody can see what already exists
Work is spread across docs, repos, tickets, chat threads and slide decks. If finding the existing churn analysis takes longer than starting a new one, people start a new one. Search is only as good as the names people gave things.
Decisions live in private places
A lot of real decisions happen in DMs, hallway calls and meetings that leave no record. Six weeks later, someone reopens a question that was already settled because there is nothing to find. A decision log fixes this more cheaply than any tool.
AI chats are private by default
AI has made it faster to produce a first draft, which means it is also faster to produce a duplicate one. Much of that AI work happens in personal chats that nobody else can see. Gartner found that 88% of employees with access to enterprise AI also use personal AI tools for work. When a colleague works out a great prompt for customer interview summaries, the rest of the team never sees it. See shadow AI at work for more on why this happens.
Tool sprawl splits the record
Context is also spread across many apps. BetterCloud reports the average company now runs 27 AI-powered SaaS apps. Each app holds a slice of the record, and no one has the full picture.
People change roles
When someone moves teams or leaves, their in-flight work and the reasons behind it often go with them. Their replacement starts over.
How to spot duplicate work early
Duplicate work is easiest to stop before it starts. Watch for these signals:
Signal | What it usually means | What to do |
|---|---|---|
"I couldn't find anything, so I started fresh" | Existing work is not findable | Name and file work so search finds it |
Two docs with the same title and different numbers | Parallel analysis, different definitions | Agree on one definition and one owner |
A settled question comes back in a meeting | The decision was never written down | Log the decision with date and reason |
A new hire rebuilds a process | Onboarding does not point to existing work | Add the source to onboarding |
Everyone has their own version of a prompt | AI know-how is stuck in personal chats | Share the best one as a team skill |
How to prevent it: six habits
You do not need a big knowledge project. Small habits remove most duplicate work.
Search before you start. Make "what already exists?" the first step of any new piece of work. Spend ten minutes looking and ask in one public channel.
Announce work when it starts, not when it ships. A one-line note ("Starting a churn analysis for Q3, owner: Priya, due Friday") gives others the chance to say "I did that last month."
Write decisions down where people look. Date, decision, reason, owner. Keep it short so people actually do it.
Give every recurring artifact an owner. The pricing doc, the competitor comparison, the metrics definitions. One owner per artifact means one version.
Share what works with AI. When someone writes a prompt that reliably produces good release notes or interview summaries, turn it into something the team can reuse. Here is how to turn one person's prompt into a team skill.
Close the loop on handovers. When someone changes roles, list their open work and where each piece lives.
A worked example: the pricing page
Say your team decides to drop the monthly plan and sell annual only. The decision happens on a call between the CEO and the head of sales.
Without a shared record, here is what follows. Marketing hears about it two weeks later and asks ChatGPT to draft a new pricing page from the old one. Support writes their own FAQ in Claude. Sales updates the deck by hand. Three people do overlapping work, and each version has slightly different terms.
With a shared record, the decision goes in the log the same day with the reason ("annual retention is higher and support load is lower"). Marketing, support and sales each start from that entry. Their AI tools get the same facts. The work still gets done three times, but each person writes their part instead of rebuilding the whole picture.
Where AI helps and where it hurts
AI is good at finding and summarizing what exists, if it can see it. Ask an assistant connected to your docs "has anyone analyzed churn this year?" and it can answer in seconds.
AI hurts when each person's assistant only knows that person's chats. Then everyone gets a fast, confident first draft built on partial context, and duplicates multiply. The fix is to give every tool the same shared context about what the team has already done and decided.
FAQ
How much time does duplicate work waste?
There is no single reliable number for duplicate work alone. Atlassian's research found that 50% of knowledge workers have unknowingly worked on duplicate projects and that 25% of the workweek goes to searching for information. Both point to the same root cause: work that is hard to find.
Is some duplication healthy?
Yes, when it is a choice. Two teams prototyping different approaches to the same problem can be a good bet. The problem is accidental duplication, where nobody knew the other work existed.
What is the cheapest fix?
A habit of announcing work when it starts, plus a simple decision log. Neither needs a new tool.
Does using AI make duplicate work worse?
It can. AI makes it cheap to start from scratch, and AI chats are usually private. If your team's AI tools cannot see what already exists, they will happily help everyone build their own version.
Make work visible
Duplicate work is a visibility problem. People repeat work when they cannot see what exists, why it was decided, or who owns it. Make work easy to find, write decisions down, share what works with AI, and give every tool the same context. Much of the duplication goes away once the team and its tools can see the same record.
We're building Shane to keep a team's decisions and work in one shared place that every teammate and AI tool can search before starting something new. If that's a problem your team has, you can join the waitlist at tryshane.ai.