Put every request in one list, score each for value and effort, and work on only a few at a time. Say no, or "not now", to the rest, with a reason. A backlog is a "might-do" list, not a to-do list (Salesforce Admins, 22-09-2025).
Clientell AI, the AI Salesforce admin, reads your org first and drafts routine changes such as field and flow edits. It asks before it changes anything in production. See the AI Salesforce admin page, or run the free, read-only health check to see what is piling up in your org.
Why does the admin backlog never shrink?
Requests come from everywhere: users, stakeholders and leadership (Salesforce Admins, 22-09-2025). Each one feels urgent to the person who sent it. One consultant put it this way: "One Salesforce admin team. Five hundred open requests. Every one of them is somebody's top priority." (LinkedIn, Esther Widmann, 16-07-2026).
There is a second cause. Old automation and messy fields create new tickets. In the Salesforce Ben 2026 admin survey, 56.3% of admins named technical debt as their biggest challenge (Salesforce Ben, 03-07-2026). Debt feeds the backlog, and the backlog stops you from paying it down.
So the fix is not to work faster. The fix is to control what comes in, what you pick, and how many things you start.
What are the exact steps, in order?
- Collect everything in one place.
- Give every request the same intake form.
- Clean the list: merge repeats, close dead items.
- Score each item.
- Set a limit on work in progress.
- Pick what to automate.
- Say no to the rest, and publish the list.
- Review every week.
The next sections cover each step.
How do I set up intake so requests stop arriving as chat messages?
Pick one home for the list. The Salesforce Admins blog suggests a Google Sheet with the columns Request, Impact, Effort, Status and Notes, a Trello board, or a custom Salesforce object for backlog items (Salesforce Admins, 22-09-2025). Use whichever your team will really open.
Then make the form short. Every request answers five questions:
| Question | Why you ask it |
|---|---|
| What are you trying to do? | You get the goal, not the solution they guessed. |
| Who is affected, and how many people? | This sizes the value. |
| What happens if we do nothing? | Separates real pain from preference. |
| What object, field or flow does it touch? | Sizes the effort and the risk. |
| Is there a date, and why? | Real deadlines are rare. Ask for the reason. |
Ask for the specific outcome. "Create a new fundraising homepage" is vague. "Design a new Lightning page so gift officers can see donation history" is something you can size (Salesforce Admins, 19-03-2026).
How do I size and prioritize requests?
Use two scores from 1 to 5: value and effort. The Salesforce Admins blog points to value and effort as the usual pair (Salesforce Admins, 19-03-2026). Divide value by effort. Work from the top.
Here is a worked example. The records are made up for the example.
| Request | Value (1-5) | Effort (1-5) | Score | Decision |
|---|---|---|---|---|
Make Opportunity.NextStep required at Proposal stage | 4 | 1 | 4.0 | Do this week |
| Fix the lead assignment flow that skips one region | 5 | 3 | 1.7 | Plan |
| New dashboard for one manager's weekly view | 2 | 2 | 1.0 | Ask for a clearer goal |
| Rebuild the quote process on a custom object | 4 | 5 | 0.8 | Needs a project plan |
| Change a picklist color | 1 | 1 | 1.0 | Batch with other small items |
Two checks keep the scores honest. First, add risk to effort when a change touches many objects or old automation. Second, let value come from the number of people and the cost of doing nothing, not from the title of the requester.
Do not score everything on day one. Score the top 20 items. The rest wait.
How many things should I work on at once?
Few. Set a limit on work in progress, for example three open items per admin. Start nothing new until one finishes. Half-done work hurts you more than a long list does: users see no change, and you carry the context of every open item.
Keep a small slice for fires. Production bugs skip the queue. Everything else goes through it.
What should I automate, and what should I say no to?
Sort the list by type, not only by score.
| Type | Examples | What to do |
|---|---|---|
| Routine and well defined | Add a field, change a validation rule, fix a page layout, merge duplicates | Automate or hand to an AI admin tool with approval |
| Needs a conversation | "Make reporting better" | Ask for the goal first |
| Big and strategic | A new quote process, a data model change | Move to a project, outside the backlog |
| Low value | Cosmetic changes nobody uses | Say no |
A clear no has three parts: the answer, the reason, and what would change it. For example: "Not now. This affects 2 people and the fix takes a week. If it blocks a sale, tell me and I will move it up." Save the no's on the list with a status of Declined, so the same request does not come back as new.
Some backlog items are also debt, such as unused fields, old Workflow Rules and duplicate flows. Put a fixed share of each week toward them. Our org cleanup post and the admin ratio guide can help you size this.
How do I do this safely in production, with a sandbox, an approval and a rollback?
Fast clearing creates risk. A bad change in production costs more than a slow week.
- Make each change in a sandbox first.
- Show the change to the requester before it goes live.
- Keep a note of the old value or version, so you can undo it.
- Ship small batches, not one huge release.
Clientell AI asks before it changes anything in production. It can test in a sandbox first, and Platform deployments have rollback. See security and privacy.
What can go wrong, and how do I undo it?
| Problem | Cause | Fix |
|---|---|---|
| The list is still huge after a month | No one closed dead items | Close anything older than your cutoff with a note, and let people reopen it |
| People bypass the form | The form is long, or the answer is slow | Cut it to five questions and reply within a day |
| A fast fix broke something | It skipped the sandbox | Roll back to the saved version, then redo it in the sandbox |
| Everything scores 5 | Value comes from opinion | Use the number of people and the cost of doing nothing |
| Stakeholders feel ignored | They cannot see the list | Share the ranked list and the reasons for each no |
Try it in Clientell AI: "Scan my org in read-only mode and give me a ranked list of the ten riskiest items, with the change you would make for each." Start the 14-day trial, no card.
Prefer to look first? Run the free, read-only health check. Clientell MCP is $249 a month and Clientell Platform is $500 a month. See pricing.
FAQ: common questions about the Salesforce admin backlog
What is a Salesforce admin backlog?
It is a running list of work your team might do: enhancement requests, process tweaks, bugs and ideas. It is not a to-do list (Salesforce Admins, 22-09-2025).
Where should I keep it?
A spreadsheet, a Trello board or a custom Salesforce object all work. Pick the one your requesters will use.
How do I say no without hurting trust?
Give the answer, the reason and what would change it. Keep the list visible so people see the order.
Will AI replace the admin who owns the backlog?
No. Someone still sets priorities and approves changes. AI can take the routine items. See will AI replace Salesforce admins.
Can Claude Code help clear it?
It can help with some tasks, with limits. See Claude Code for Salesforce admins.
Sources
- Why Every Admin Needs a Backlog, Salesforce Admins, 22-09-2025
- How Can Agentforce Help Manage a Salesforce Backlog?, Salesforce Admins, 19-03-2026
- The State of the Salesforce Admin Role in 2026, Salesforce Ben, 03-07-2026
- Esther Widmann, LinkedIn, 16-07-2026
Last updated: October 9, 2026
