Business & Tech · The Lithos Blog
In March 2026, Atlassian cut roughly 10% of its workforce to self-fund AI investment. You don't need to cut anyone — but the message for every business owner is clear: the way routine work gets done is changing. Here are six practical workflows to start with.
The Wake-Up Call
On March 11, 2026, Atlassian — the software company behind Jira and Confluence — announced it was eliminating roughly 1,600 roles, about 10% of its workforce. Not because revenue collapsed. The stated purpose: to self-fund increased investment in AI and enterprise sales. And Atlassian isn't alone — Block announced roughly 4,000 cuts around the same period as part of its own AI-driven restructuring.
Now, here is the honest, mentor-level read: analysts still debate how much of this is genuinely AI and how much is ordinary cost pressure wearing an AI label. Both can be true. Either way, the direction is unmistakable — routine process work is being reassigned to software, and the companies moving first are redirecting the savings into more capability.
You run a small business, not a 16,000-person software company. You don't need to cut anyone. But you do need to understand what just became possible — because it's available to you at small-business prices.
Per Gartner's forecast, task-specific AI agents will be embedded in 40% of enterprise applications by the end of 2026 — up from less than 5% in 2025. That's not gradual change. That's a step-change in a single year.
AI 1.0 vs. AI 2.0
For two years, most of us "used AI" by typing a question, getting a paragraph, and pasting it into an email. That was AI 1.0. The shift now underway is the agent — and the difference matters for how you run your books.
You type a question. It gives you a paragraph. You edit and send. You're still doing the work — the AI is a research assistant with better grammar.
You give it a goal and a boundary. It works across your inbox, documents, and spreadsheets to complete a multi-step task — then reports back for your review.
An agent doesn't just answer — it does. What that looks like in a finance workflow is less science fiction and more of this:
Every weekday at 7:00 am —
read the new invoices and receipts in the bookkeeping inbox,
categorize each one, flag anything missing a PO or W-9,
and draft the follow-up emails for my review before 8:00.
Notice the last three words: for my review. In a well-designed small-business workflow — especially in regulated industries — the agent does the boring 80%, and a human makes the judgment call at the end. That's the standard every workflow below follows.
The Playbook
You don't need a five-figure custom build. You need focused workflows that attack the highest-cost, lowest-leverage work in your week. These six are illustrative patterns we see working — your mileage depends on your tools, volume, and setup:
An agent reads incoming mail, classifies it — reply now, reply later, route, archive — and drafts responses for the first two categories, in your tone. You review and send. For most operators this is the single largest weekly time reclaim, and the cheapest to pilot.
Every invoice, receipt, 1099, and W-2 that arrives gets read, categorized, and checked against what's expected — with missing or mismatched documents flagged and a polite request drafted automatically. Picture a boutique CPA workflow where client documents are classified on arrival instead of piling up for a Friday sort.
Before any sales or client call, an agent assembles a one-page brief — public company news, filings or listings, mutual context, and your own past notes. You walk in prepared without spending the prior evening digging.
Record one long-form piece — a podcast, a webinar, a client explainer — and an agent drafts the short clips, article versions, and social posts in your voice, queued for your review. One hour of recording becomes weeks of educational content.
For regulated businesses — insurance, financial services, real estate — an agent watches the CRM and flags any client file missing a required form, an expired ID, or a pending signature. Problems get caught before a compliance review finds them, not after.
An agent watches the public data that matters to your niche — listings, filings, local business registrations — and drafts timely, personalized outreach for your review. The human relationship still closes; the software just makes sure you show up at the right moment.
Notice what's happening across all six: the work isn't being eliminated — it's being reassigned to software. The owner's job shifts to designing the system, editing the output, and having the human conversation at the end. That's not a smaller job. It's a better one.
The Real Question
Roles built entirely around process work — receive input, look something up, produce output, with no relationship layer and no judgment call — are exactly what agents now do quickly and cheaply. Atlassian's cuts fell hardest on process-heavy functions. If part of your business runs on that pattern, the honest move is to start redesigning it now, on your terms.
Building the relationship, designing the strategy, holding a client's hand through a difficult moment, making the judgment call — agents don't replicate that. They clear the routine 80% so you can be fully human at the 20% that actually matters. That's where your leverage lives, and it's exactly the part of the work that grows a firm.
How to Start
The most common mistake is starting with a large custom build. Don't. Pick one workflow you do every single week — and dislike doing. Build for that one thing, watch it work, then iterate.
Be specific: "Three hours every Monday pulling data into a spreadsheet"
Ask an AI assistant to design the agent and list the tools it would need
Start with a low-cost tool and a human-review step on every output
Hours reclaimed × what you'd do with them. If the answer is selling, building, or family — go
And this is no longer experimental plumbing. The infrastructure has gone mainstream — Microsoft's Agent 365 reached general availability in May 2026 as a control plane for governing agents across business systems, and every major platform is shipping its equivalent. The tooling question is largely solved. The remaining question is which workflow you hand over first.
The Long View
A few humans on top, making judgment calls and building relationships. A set of agents underneath, handling the routine. The owners who get there first won't win because they're smarter — they'll win because they started earlier and compounded the reclaimed time.
Large companies are restructuring around agents and redirecting the savings into more capability — the pattern of 2026.
You don't have to cut anyone. You do have to start building. Every week of waiting is a week the gap moves — one direction or the other.
Don't be the business still paying full price for routine work software now does. Be the one collecting the difference — and reinvesting it in the human side.
These resources support the facts and research referenced throughout this article.
Coverage of the March 2026 restructuring: roughly 10% of the workforce cut to self-fund AI investment, part of a wider industry pattern.
thenextweb.comThe original press release: task-specific AI agents embedded in 40% of enterprise applications by end of 2026, up from under 5% in 2025.
gartner.comThe governance layer arrives: Microsoft's control plane for deploying and managing AI agents across business systems reached general availability in May 2026.
microsoft.com
Dr. Saha works with families, business owners, and aspiring entrepreneurs on financial education, wealth strategy, real estate, and entrepreneurship — helping people build stronger foundations and make confident decisions. Meet the author →
Next step
Whether you're ready to build or just want to understand what's possible for your specific business, let's talk. We work with small business owners, insurance professionals, and financial-services operators — and we use these same workflows in our own practice. Education first, always.