What to Outsource in the AI Era — Three Buckets

A friend recently asked me: "AI is so capable now — do we still need software outsourcing companies?"

The answer: more than before, but what you outsource is completely different.

The old model = hire a team of engineers to write code. As AI keeps driving the cost of code down, that logic broke. For the same money, what you should get isn't a feature count — it's better judgment + AI leverage + a fully redesigned process.

Three buckets to keep separate:

1. Don't outsource it — what you can do yourself with AI
2. Do outsource it — what doesn't pencil out in-house, but where the data matters
3. Process-consulting outsourcing — if you want to gradually hand off day-to-day work to agents

1. Don't Outsource It: Just Use AI Yourself

Many things that used to require a vendor can now be handled internally with AI tools:

• Static website pages: landing pages, product info, campaign pages — AI generates them, marketing owns the workflow
• Anything you can change with a prompt: copy edits, image swaps, layout tweaks, color changes. What used to need an engineer queue is now one sentence
• One-off small systems: event registration, internal surveys, short-term intake forms — disposable, AI + a template gets you there
• First draft of a manual / FAQ: AI writes the draft, a human polishes
• Small bugs with clear error messages: throw it at AI, usually solved in 10 minutes
• Small, non-critical data conversion: Excel to CSV, field mapping. But important data (customer, finance, ERP master) always needs human review

The precondition is your company identifies an AI-savvy employee — doesn't have to be an engineer, just someone who can use AI to ship the things above. Without this person, every "do it yourself" option stalls and you end up back at outsourcing.

Going wider: this isn't "find one person and you're done" — every department needs someone who applies AI to their workflow: marketing, sales, customer support, HR. AI isn't an IT-only thing; it's everyone's tool.

2. Do Outsource It: Systems That Run Daily and Accumulate Data

The remaining 90% of budget should go to things AI can't do. From the data angle, the boss's job is: identify which data in your company will be valuable later — those are the systems worth outsourcing and customizing.

Data is what you feed an AI agent: HR attrition data trains retention prediction; CRM conversation history feeds a customer-service agent; expense flows feed anomaly detection. Without data, an agent is just an empty shell — so "where your data lives, who owns it" matters more than "what the system looks like".

From a data angle, all three old options are broken:

• In-house team: own your data ✓, but $100-200K/year per senior engineer doesn't pencil out
• Full outsource: own your data ✓, but small projects get over-quoted
• Generic SaaS: cheap, but data locked in their system — exporting, API access, feeding it to AI all friction. Your most valuable asset in the AI era is your data, not someone else's system ✗

Most SMBs end up "muddling along with Excel + Google Forms" — data scattered across personal hard drives, even they can't use it.

AI changes the economics: outsourcing is faster (and cheaper) with AI assist, maintenance is lower (routine handled by AI, humans handle exceptions), low-code makes assembly fast. Core systems stay in-house, operational systems get outsourced and customized — but the data always stays with you. That's the AI-era division of labor.

3. Process-Consulting Outsourcing: Agentify Your Workforce, Gradually

This is what bosses should think about most. The real game changer in the AI era is: gradually replacing human-done work with agents. Compared with building yet another website or custom system, the long-term ROI is an order of magnitude higher.

Why It's the Highest Long-Term ROI

Because agentifying labor is a compounding investment — every process you automate is a permanent capacity gain you don't keep paying for. Building a website or CRM is one-shot; agentifying is long-term compounding.

Process redesign used to be hard — you needed a full team (consultants + PMs + engineers) to make it happen, and SMBs couldn't afford it. So they made do with Excel + manual work.

The AI era is different: AI agents + off-the-shelf SaaS + low-code shrink the team and compress the timeline. SMBs can finally afford it.

AI is also collapsing the wall between "consulting" and "IT outsourcing" — one used to make PowerPoints, the other wrote code, with handoff translation losses in between. Now one firm can take it from discovery → design → build → ship. What you're really buying is a product manager who understands business processes — what most SMBs lack isn't an engineer (AI fills that), or a strategy consultant (PowerPoints don't ship). It's this person.

How it differs from regular software outsourcing

	Regular software outsourcing	Process-consulting outsourcing
What you buy	A system (website, CRM, backend)	Process redesign + agentification
Start with	You give specs, vendor builds	Consultant runs discovery first, then decides what to build
Right for	You know what to build	You know there's a process problem, don't know what to build
AI-era ROI	Medium (scope shrinks)	High (things you couldn't do before are unlocked)

Four things to look for in a process-consulting partner

#	Ask	What you want to hear
1	How do you run discovery?	A concrete workflow, not just handing you a questionnaire
2	How do you make processes agent-driven?	Specifically: how agent and human hand off, how permissions are set, who decides on errors
3	What AI tools do you use? Can we take it over later?	Mainstream stable tools, not lock-in
4	After launch, how do you support iteration?	Full documentation, transition support, long-term partnership

Warning: don't try to fit a process problem into a regular software outsourcing contract — they'll build you a system "with very clear specs that solves nothing real".

Closing

AI-era outsourcing isn't a cost-cutting tool — it's a lever to amplify what your organization can do.

The boss's mental order:

1. Don't outsource what you don't have to (use AI internally)
2. Build an AI team inside the company: not the IT department, every department needs someone who plugs AI into their workflow
3. Agentify your workforce, gradually: find a process-consulting partner who can integrate into and improve the organization — this is the real opportunity in the AI era