The biggest challenges Salesforce admins face in 2026 are not necessarily new. Technical debt, the standard-vs-custom decision, data quality, an expanding role – these were already the job a year ago.
So why does the role feel heavier? Because these challenges share a root cause – complexity that grows faster than it gets cleaned up – and this year something finally started charging for it.
AI rarely works well on top of a messy org, so the pile of "we'll fix it later" work that admins carried quietly for years turned into the magnifying glass that made its cost impossible to ignore.
The annual SF Ben Salesforce Admin Survey gives us a way to see what actually shifted from last year and why the quiet challenges are the ones that will define the role next.
The challenges themselves barely changed between 2025 and 2026. The real movement is somewhere the ranking doesn't show it: in what those challenges now cost.
AI is what raised the price. Adoption crossed into the mainstream – Agentforce usage doubled to 30%, and the share of admins saying AI makes them more productive climbed from 62.3% to 71%, the biggest single move in the whole dataset. But confidence didn't follow, and security is now the skill admins feel least sure of.
So AI didn't rewrite the list – it made the accumulated mess underneath it expensive, re-casting old problems like technical debt, data quality, and governance as prerequisites for AI work rather than standalone chores.
The table below tracks the numbers from the 2025 Admin Survey (793 respondents) to 2026 (1,100+ across 72 countries).
The five challenges that follow explain why each one still matters – and what changed since last year.
Technical debt remained the most commonly cited challenge, named by 56.3% of admins. What the raw ranking hides is how few orgs feel on top of it: only 2% describe their environment as clean and well-maintained, while 31% report levels high enough to slow day-to-day delivery.
It is also unevenly distributed by experience – 13% of entry-level admins feel confident handling it, against 44% of advanced admins – which means the people most likely to inherit a messy org are often the least equipped.
.png)
The year-on-year story is subtle. The share naming tech debt as their biggest problem actually fell from around two-thirds in 2025.
That is not a cleanup victory. It is that other challenges – AI, security, integrations – climbed the list, so tech debt now competes for attention it used to own outright.
And there is already more work coming. With the Winter ’27 release window open, admins need to make room for another round of changes – which is harder when the existing backlog is still competing for attention.
That is the real cost of debt. It serves as the accumulated tax on every small change.
2025 → 2026 verdict: No longer alone at the top. AI both adds to that debt and exposes it – why, keep reading further.
The way out starts with visibility – debt captured in a single backlog and ranked by impact: what slows delivery, what's still unresolved, what's cosmetic.
Then stop making the backlog bigger: CI/CD quality gates on critical vulnerabilities, new-code coverage, duplication, and complexity, set to fail a build when agreed thresholds regress. The fastest debt reduction isn't “cleanup” – it's preventing every release from adding more.
Balancing native functionality against custom builds was cited by 39.4% of admins, and it earns its place because it is where tomorrow's technical debt is quietly created today. Every requirement sits on a spectrum: standard tools are fast to ship but sometimes too rigid, while customization fits precisely but has to be maintained for years by people who may not have built it.
What changed in 2026 is the stakes of getting it wrong. With AI-assisted building lowering the effort of creating a custom Flow or object, the constraint is no longer how fast you can build – it is whether anyone will understand the build later.
The most durable admins evaluate not just what solves the problem now, but what will still be supportable after two or three release cycles.
2025 → 2026 verdict: Steady in ranking, higher in consequence. Now that AI makes building fast, the harder skill is knowing when not to build – every shortcut today is someone's maintenance burden tomorrow.
A custom build's real cost shows up years later, in everything it touches. What contains it is seeing that blast radius up front – reading the org to know what a change connects to before it's built, so the decision to customize is made with its downstream cost transparent.
The clearest theme running through the 2026 data is not any one task – it is capacity. Some 59% of admins say Salesforce is getting more complex, 53% say too much is expected of them, and 43% work in admin-only teams with no dedicated developer.
Nearly one in five is a solo admin running an entire org alone. The role now routinely spans automation strategy, integrations, data quality, reporting, and elements of security governance, all at once. Consultant Vuk Zdinjak captured the trajectory in a line that circulated widely:
The administrator of 2016 has become the business analyst of 2026.
As AI lowers the barrier to technical work, judgment – knowing which problems are worth solving and which requests to decline – becomes the scarcer skill. That reframes "too much is expected" from a complaint into a career signal: the expansion is real, and the admins who thrive are the ones who move up the value chain rather than trying to do everything.
2025 → 2026 verdict: Flat on paper, sharper in practice. The expectations gap didn't widen, but AI raised what "keeping up" means.
The relief isn't working harder across more areas – it's work that starts without someone having to remember it. Introduce a governed intake process to protect recurring capacity for maintenance, security, and technical debt.
Check the skills the work needs against what the team covers now, then close the gaps that count – with training, somewhere to escalate, and external specialist support to lean on.
Every survey year, some version of "keeping data clean" ranks among the most time-consuming parts of the job, and 2026 is no exception. Duplicate records, incomplete entries, and inconsistent free-text values are individually trivial and collectively exhausting.
Certified admin Ryan Kawashima described the texture of it well to Salesforce Ben:
At my last place of employment, the task that ruined my day was merging duplicates. It wasn’t hard, but it consistently took time out of my day that could’ve been spent on strategic vs. day-to-day work.
The reason this challenge refuses to leave the list is the same reason it is easy to ignore: no single duplicate is a crisis, so the cleanup work never becomes anyone's priority until it blocks something bigger.
In 2026, that something bigger is AI. Grounding features on a dataset full of duplicates and stale records tends to produce confidently wrong results, not neutral ones. Salesforce's built-in duplicate management tools, validation rules, and required fields remain the first line of defense, and they matter more now that data quality has an audience beyond the reporting team.
2025 → 2026 verdict: Unchanged in difficulty, promoted in importance. The routine is now a dependency.
The lever is spotting the problem in the org itself, not waiting for it to surface downstream. Baseline and segment first: profile Accounts, Contacts, and Leads with native matching rules, measure exact and fuzzy duplicates, assign data owners, and define survivorship rules before merging records.
Native tools handle the baseline, but at scale most teams reach for a dedicated deduplication tool – and the ones admins trust tend to be the ones with a track record on real orgs.
.png)
This one rarely appears as a line item on challenge lists, but it sits underneath half of them. AI adoption among admins has crossed into the mainstream: 44% now use it daily or regularly, daily use nearly doubled year on year, and 71% say it makes them more productive.
Yet confidence hasn't followed – Agentforce is still one of the areas admins feel least sure of, and 54% are unaware of the Shared Responsibility Model that defines who secures the data feeding those tools.
The pattern that emerges across practitioner accounts is consistent: AI initiatives succeed or fail on the foundations admins already own. Salesforce MVP Vicki Moritz-Henry put it plainly in conversation with Salesforce Ben – orgs cannot be ready for Agentforce or any AI toolset while they carry dirty data and unresolved technical debt, and admins are the ones pulled in to fix the ground before the AI work can start.
2025 → 2026 verdict: The challenge today is not "learn a new AI feature." It is recognizing that AI concentrates existing responsibilities rather than removing them – and getting the data, security, and governance house in order first.
The gap isn't the agents – those exist. What's missing is the harness around them. Establish a shared agent harness for org context, permissions, tool access, approvals, monitoring, and rollback.
Standardize how agents understand Salesforce metadata, execute actions, log decisions, and escalate exceptions. Build reusable agent infrastructure first, then add bounded use cases without recreating governance each time.
Everything above is the look backward. The more useful question is where it leads – because the shift that started in 2026 doesn't stop there. Here's what the next year is likely to demand.
Every challenge above points the same direction, so the honest forecast is simple: 2027 won't add new problems – it will start grading the old ones.
For years admins could carry technical debt, dirty data, and undocumented builds as background noise. AI removes that luxury, because an agent can only be as reliable as the org underneath it. The messy foundation that was probably survivable in 2025 becomes the thing that stalls the AI project in 2027.
That turns a soft skill into a hard one. The scarce ability is no longer building faster – AI already builds fast. It's judgment: knowing which requests to refuse, which customization will still make sense in three release cycles, which cleanup unblocks everything downstream.
It's the same shift the survey hints at when admins move toward analyst and architect work – the value migrating from doing the task to deciding whether the task should exist. Alec Green, IWU National & Global’s Associate Director of Salesforce Development, has exactly described that:
The skill gap to building something is shrinking in some ways. The knowledge of whether you should build it, and how to build it right, is becoming more valuable than ever.
Technical debt, data quality, and the standard-vs-custom balance are the same challenges admins faced last year, but each now carries a second layer: they are prerequisites for the AI work that the role landscape increasingly expects.
More admins use it and more say AI helps, but confidence didn't follow and security is now their weakest spot. So the real shift isn't that artificial intelligence solved anything or broke anything. It's that the gap between using AI and being ready for it widened – and that gap closes through the same foundations that have topped this list for two years running.
The admins who come out ahead are not the ones with the longest tenure. They are the ones who treat the foundational work – clean data, documented changes, deliberate customization – not as chores to squeeze in, but as the thing that makes everything else, including AI, possible.
What is the biggest challenge for Salesforce admins in 2026?
Technical debt, named the most challenging task by 56.3% of admins. Only 2% describe their org as clean and well-maintained, and 31% carry enough debt to slow daily delivery.
Did admin challenges change much from 2025 to 2026?
Technical debt fell from 67% to 56.3% but stayed #1, "too much is expected" held flat at 53%, and complexity eased only slightly. The real change wasn't the problems; it was AI raising the cost of ignoring them.
How is AI changing the admin role?
AI adoption is the one metric that genuinely moved – Agentforce usage doubled to 30% and 71% of admins say AI makes them more productive, up from 62.3%. But confidence didn't follow, and security is now the skill admins feel least sure of. AI concentrates the responsibilities admins already carry rather than removing them.
Why does technical debt keep topping the list?
Because the work is continuous and rarely urgent. No single duplicate, unused field, or undocumented Flow is a crisis, so cleanup gets deprioritized until it blocks something bigger – and in 2026, that something is usually an AI project.
What should admins focus on going into 2027?
The foundations: clean data, documented changes, sane permissions, and deliberate customization. It's the same unglamorous work that's topped this list for two years – and it's exactly what makes AI initiatives succeed instead of stall.
.png)
.png)
.png)