Menu
The company shares a brand. Delivery still runs like several companies.
Scoping looks different depending on which unit picked up the work. Briefs arrive in different shapes. Status means one thing in one room and something quieter in another. Forecasting eats time because nobody trusts a single picture. Acquisitions got standardized where they could. A lot of differences remained.
Someone eventually says the obvious next sentence: standardize and automate.
Automate first and you encode the differences. Standardize first and you decide which differences deserve to survive.
That isn't mostly an AI problem. It's stitched delivery ops: shared management work running on uneven habits across units.
Across multi-unit agencies and similar mid-market operators, a familiar set of jobs shows up in every unit even when the craft work differs:
Those jobs are the spine of how work enters, moves, and gets reported. When each unit invents its own version, leaders can't see across the portfolio, people relearn the company at every handoff, and automation has nothing stable to attach to.
Craft can stay different. Intake and status discipline usually shouldn't.
We keep seeing the same picture without needing a named account to make the point.
Several business units, each still working pretty differently after acquisitions. The common management work (scoping, briefing, forecasting, project management, status) is named as the place to standardize and automate. Standardization happened where it could; differences remain. Maturity varies widely by unit. Someone close to the work is trying to move fast on agent-shaped ops while still missing a clear operating picture of how delivery actually runs today.
That is one problem, not a tooling shortage. You don't need an agent on status first. You need the shared delivery jobs written the same way where sameness matters.
Start with the jobs that create a single company view and a survivable handoff.
Usually the same:
Often allowed to differ:
If everything is forced identical, units resist and shadow process returns. If nothing shared is identical, automation just speeds up confusion.
Forecasting is often the loud time sink because it sits on top of uneven intake and status.
If units scope differently and report status differently, forecasting becomes manual translation. People rebuild the truth every cycle instead of reading it.
Treat forecasting as a dependent system. Stabilize scoping and status definitions first. Then forecasting has something trustworthy to read. Automating forecast spreadsheets while intake still means five things will create a faster wrong picture.
Before any agent-shaped build on delivery ops, write down:
If those lines aren't written, an agent on SOWs or status will inherit every unit's private language. You will not have automated delivery. You will have automated disagreement.
Custom builds belong after that documentation exists. The operating work comes first. The build is how you encode a path the units already agreed to run.
Which delivery work should be the same across units, and what can stay different?
Make sameness where portfolio view and handoffs break: scope readiness, brief structure, status meanings, and where artifacts live. Let craft methods and some client ritual differ if they don't break shared definitions. Force everything identical and shadow process returns. Share nothing and automation has no stable target.
Where does forecasting fit?
After scoping and status definitions stabilize. Forecasting becomes manual translation when units mean different things by intake and progress. Automating forecast views first usually speeds up a wrong picture. Stabilize the inputs, then forecast from them.
What gets documented before an agent touches status or SOWs?
Shared definitions, owners per stage, allowed local exceptions, systems of record, and human review points for contractual or client-facing output. Without that, an agent inherits every unit's private language.
If the pattern on this page sounds familiar, start with the free AI Operations Reality Check. One everyday workflow. No pitch deck.
If you want a bounded diagnostic and a written next-step memo, the AI Operations Review is the paid step after that ($997, not an open-ended contract).
Susan Westwater is the CEO and Co-Founder of Pragmatic Digital. She helps mid-market and PE-backed teams move from scattered AI pilots to governed, measurable workflows that actually deliver operating leverage. With 25+ years in CX and brand leadership at Leo Burnett and Ricoh USA, Susan specializes in turning AI ambition into repeatable systems that protect brand voice and reduce revision cycles. She is co-author of Voice Strategy and Voice Marketing.
Weekly insights on AI adoption, workflow, and what's actually working in mid-market organizations.