How AI Changes the Salesforce Consulting Model

A traditional Salesforce consulting engagement bills you for hours. The hours are real — somebody has to architect the solution, configure the org, write the Apex, run the test cycles, document the result. Each of those steps has a cost, and the agency's margin is the gap between what you pay and what they pay the engineers.

AI is collapsing the cost of two of those four steps. Architecture is still hard. Test cycles still need real-world validation. But configuration grunt-work and Apex/test/doc generation are now 60-80% faster with the right workflow. And nobody's billable rate has fallen yet to reflect that.

The gap is a temporary arbitrage. The consulting model adjusts to it eventually — the question is whether your supplier adjusts now, in two years, or after you switch.

What's Genuinely Cheaper

A few things AI has made measurably faster in my workflow:

Boilerplate Apex. Trigger handlers, test classes, batch job scaffolds, REST endpoints with standard auth patterns. AI generates 80% of this in minutes. The 20% that needs editing matches the project's voice and patterns; the structure is free.

Bulk refactoring. Renaming a method across 50 classes. Adding with sharing to every class missing it. Converting System.debug calls to a logging utility. Tasks that took hours now take minutes, with a clean diff to review.

Documentation. Class headers, method docs, README files for packages, runbook entries. AI does the structure. You add the institutional context.

Test data factories. Generating realistic test data with all the relationships and required fields takes real time. AI can scaffold a factory that handles 80% of the cases; the remaining 20% comes from production-edge knowledge.

Field mapping for migrations. Mapping 200 fields from a legacy CRM to Salesforce by hand takes a day. AI does a first-pass mapping in 5 minutes; you review and override the cases it got wrong.

If your consulting partner bills $200/hour for these tasks, they're billing you for work that now takes a fraction of the time. The question isn't whether they're using AI — most are. The question is whether their pricing reflects it.

What's Genuinely More Valuable

The flip side: some skills are more valuable now, not less.

Architecture decisions. AI can generate code that fits patterns. It can't tell you which pattern fits your specific business context. The cost of bad architecture compounds — and AI accelerates that compounding because more code can be written faster on top of bad foundations.

Knowing what to build. Translating a vague business requirement into a concrete Salesforce design is judgment work. AI helps execute the design but can't make the design. Senior engineers who can do this well are scarcer, not less so.

Production debugging. When something breaks at 11pm and the integration is down, you need someone who can read the actual logs, understand the actual data shape, and reason about a system in motion. AI doesn't have access to your org's runtime state. The person on the other end of the page does.

Reading what's actually in the org. Org-level intuition — that trigger fires before that flow because of the order custom metadata defines it; this picklist value breaks integration X; this user has a permission set that's silently overriding the profile — comes from operating in real orgs at scale. AI can't substitute for this, and the people who have it are worth more, not less.

The Agency Model is Under Pressure

The standard agency model is built on a labour pyramid. Senior architects design solutions. Mid-level engineers configure orgs. Junior engineers write tests and docs. The agency makes margin on each layer, and the pyramid is wider at the bottom.

AI compresses the bottom of the pyramid. Most of what junior engineers do — writing test classes, generating boilerplate, drafting documentation — is what AI does well. Some of that work goes away entirely. The rest gets done by mid-level engineers who can supervise AI output instead of waiting for a junior team.

The implication: agencies that historically billed $150/hour for junior work and $250/hour for senior work can no longer maintain the margin on the junior layer. They have to either drop the rate, drop the layer, or pretend the work still requires junior hours.

I've watched several Australian agencies pretend. They send the same junior team to a project, but the junior team is now using AI to deliver the work. The agency bills the same hours. The client pays the same fee. The agency's effective margin doubles. This works until the client figures it out — and clients are figuring it out.

What the Replacement Model Looks Like

The model that actually fits the new economics is closer to what I run:

Fixed-fee, scope-bounded engagements. Health checks at $750. Migrations at $2-5k. Automation builds at $500-2k. The price reflects the deliverable, not the hours. If AI lets me deliver a $2k migration in 3 days instead of 3 weeks, the client pays the same and gets it sooner.

Senior delivery, fewer hands. Instead of 4 people across 3 levels, one senior engineer with AI tooling does the same work. The depth and judgement are higher because every line passes through someone who can think about it. The cost is lower because there's no hand-off overhead.

Async-first. Calls happen when they're the cheapest way to converge — design kickoff, mid-engagement checkpoint, final walkthrough. The rest is written briefs, recorded walkthroughs, and PRs. The traditional weekly status call is replaced by a written status update that takes 5 minutes to read.

Outcomes documented for the client to keep. Not a deck. Not a slide pack. Markdown documents the client can grep, search, and version. The deliverable survives the engagement and the consultancy isn't required to interpret it later.

This model works because the bottleneck is no longer engineer-hours. The bottleneck is the engineer's judgement applied at the right places. AI multiplies the output of judgement; it doesn't generate judgement itself.

What Clients Should Demand

If you're a Salesforce buyer in 2026:

Fixed-fee engagements. "Time and materials" is a structure that benefits the seller. Demand a fixed price. If the consultant can't quote one, they don't understand the scope well enough to bid.

A senior person doing the work. Ask whether AI is part of the workflow and who's reviewing the output. The answer should be "yes" and "the senior engineer." If the answer is "we don't really use AI" or "the junior team handles output," that's a signal — either the agency is behind, or they're billing you for inefficiency.

Working artifacts, not just slides. A health check that's a PowerPoint is a deliverable from 2015. A health check that's a markdown report you can share with your team and grep three months later is a 2026 deliverable. The same applies to architecture documents, migration plans, and project summaries.

Async delivery. If your consulting engagement requires a weekly 60-minute status call, you're paying for meeting overhead. The work should be visible in commits, PRs, and written updates. Calls happen when there's actual signal to converge on.

What's Still Worth Paying For

A few things AI hasn't changed:

Senior judgement under uncertainty. Decisions about whether to use Flows or Apex, whether to add a custom field or extend an existing one, whether to build now or wait for the platform feature next release — these require domain context AI doesn't have.

Production accountability. Someone who'll be on the call when the integration breaks. Someone who'll write the post-mortem. Someone whose reputation is tied to the long-term success of the org, not just the project deliverable.

Org-level intuition. Knowing why your specific org behaves the way it does, what the previous five teams built and abandoned, where the bodies are buried. This is local knowledge built over time. AI can read documentation; it can't substitute for time spent in the actual org.

Honest trade-off conversations. AI tends to hedge. Senior consultants tell you when an idea is bad, when scope creep is happening, when the technology is wrong for the use case. That candour is hard to find and worth paying for.

The Honest Self-Interest

I run my consulting practice this way because it's the model the market wants — and because it's the model I want to deliver. The fixed-fee, async-first, AI-augmented approach works for clients who want senior delivery without agency overhead. It also works for me because I can help more clients with the time available, doing the kind of work I find interesting, instead of running a team to amortise overhead.

The agencies that adapt will be fine. The ones that don't are running on borrowed time — they've been arbitraging the gap between AI-augmented productivity and AI-pretending billing for a couple of years now, and that arbitrage is closing.

The shortest summary for buyers: pay for judgement and accountability, not for hours. Pay for outcomes documented to survive the engagement, not for slides. And ask hard questions about how the work actually gets done — the answers tell you whether you're paying for results or for theatre.