Find where work gets stuck.
Map the process, the handoffs and the exceptions. Turn scattered information into an agreed view of what happens today.
For consulting firms and agencies
I help consulting firms and agencies untangle workflows, understand project capacity and decide where AI or automation is worth the effort.
For consulting firms and agencies with roughly 10–50 people. Based in Norway. Working internationally.
Demand / capacity · hours per week
Delay scope, move work or add capacity. Total hours alone hide the constraint.
A useful starting point is a recurring problem you can describe: work waiting for an owner, conflicting demands or reporting that takes too long.
Map the process, the handoffs and the exceptions. Turn scattered information into an agreed view of what happens today.
Compare demand with available capacity. Assess improvements against effort, risk and likely value.
Clarify ownership, simplify the workflow and specify a small pilot. Test AI assistance where it makes sense.
Two reconstructed professional examples and one fictional audit. Each is clearly labelled; no confidential employer data or claimed client savings.
A clear view of demand, usable capacity and the decisions needed when several projects depend on the same people.
At Statkraft, my work included resource coordination, project portfolio reporting and recurring resource meetings. This case illustrates how I would structure a capacity decision. The example, numbers and artefacts below were independently reconstructed; they are not an employer deliverable or a record of measured results.
A project-based team has several concurrent commitments and a shared analyst. Every project looks manageable in isolation, but the same analyst is needed in the same week. The useful question is which commitments can be supported with the capacity actually available.
Agree the planning unit and time horizon; separate committed demand from tentative requests; subtract leave and recurring operational work from available hours. Validate the demand with project owners and availability with resource owners. Missing or uncertain estimates remain visible rather than being treated as zero.
| Role | Usable hours | Demand hours | Gap |
|---|---|---|---|
| Business analyst | 36 | 48 | 12 short |
| Process specialist | 32 | 30 | 2 spare |
| Team lead | 24 | 22 | 2 spare |
Total capacity is 92 hours against 100 hours of demand, but the analyst has a 12-hour shortfall. Four spare hours in other roles do not solve that bottleneck unless the work is transferable. A portfolio average would hide the constraint.
| Option | Effect | Condition |
|---|---|---|
| Defer 8 analyst hours | Shortfall becomes 4 hours | Project owner accepts revised sequence |
| Transfer 4 analyst hours | Closes remaining gap | Split 2 + 2 across the two roles, if skills match |
| Keep all commitments | Shortfall remains 12 hours | Sponsor must resolve timing or capacity |
Validate the eight-hour deferral first. Split the remaining four hours across the process specialist and team lead, two hours each, only after checking skills, access and supervision needs. If the work is not transferable, record the unresolved four-hour gap and escalate the schedule decision. Do not silently assume overtime or interchangeable roles.
A capacity-and-demand table with named data owners, a short list of conflicts, a decision log and a repeatable update routine. A weekly update can focus on changed commitments; the decision meeting should focus on unresolved exceptions.
Before changing the process, record how often conflicts are identified late and how long it takes to assemble a reliable view. After several planning cycles, compare those measures and ask resource owners whether the model reflects reality. No improvement percentage is claimed in this case.
Structuring uncertain operational data, distinguishing a capacity problem from a prioritisation problem, and presenting practical options for management. The historic evidence supports my participation in resource coordination and reporting; the model and decisions here are reconstructed examples.
A workflow that makes ownership, required inputs and exceptions visible before anyone starts automating it.
At Statkraft, I worked on process modelling in ARIS and workshops with process owners. My earlier EY experience included business-process and internal-control work. This generic service-request example was reconstructed for the portfolio. It does not reproduce a Statkraft or EY process, client finding or internal document.
A service team receives work through email and informal messages. Requests may move to delivery before the scope is complete. The demonstration asks how to make the normal path, decision points and exceptions explicit without adding unnecessary administration.
Walk through one normal case and one exception with the process owner. Identify the trigger, required inputs, actual decisions, systems of record and handoffs. Separate observed steps from assumptions and proposed changes. Return the current-state map for validation before designing the future process.
Incomplete request: return to the requester with missing fields and an owner. Urgent request: record who authorises an exception before work starts. Scope change: pause the affected commitment until the accountable owner accepts the change. No silent bypasses or automatic approvals.
| Checkpoint | Accountable role | Evidence |
|---|---|---|
| Required inputs complete | Coordinator | Checklist in request record |
| Scope and priority agreed | Team lead | Approval and rationale |
| Delivery owner accepts | Delivery lead | Assigned owner and due date |
| Scope change accepted | Team lead | Change entry with decision |
| Work accepted and closed | Requester | Acceptance recorded |
Required-field checks, routing rules and reminders are candidates for simple rules once responsibilities are agreed. AI may help draft a request summary, but a person should approve scope and priority. The map establishes what a system must support; it is not proof that any integration has been built.
A validated current-state map, a future-state map, a process narrative, an ownership/control register, an open-questions list and implementation requirements. A simple flow diagram is sufficient for this example; formal BPMN can be used where the client needs that notation.
Ask each owner to confirm their inputs, decision and output. Test the proposed workflow with a complete request, an incomplete request, an urgent request and a scope change. Then measure missing-input rework and unowned requests against a baseline. This portfolio does not claim historical reductions.
Connecting stakeholder discovery, process documentation and control awareness. The professional background is documented in career sources; the specific workflow, register and recommendations are newly reconstructed.
An assessment for a fictional 18-person service business, with explicit assumptions, a ranked shortlist and a human-reviewed pilot.
Morrow Studio is an invented 18-person B2B service business. Every volume, time estimate, cost and projected benefit in this case is a demonstration assumption. No interviews were conducted and no system was deployed. The purpose is to show how I would evaluate operational improvements and challenge an AI business case.
Requests arrive in a shared inbox, project updates live in spreadsheets and meeting decisions are copied into task lists. The operations lead wants less chasing and clearer reporting. The assessment covers intake, weekly reporting and meeting follow-up; it excludes customer-facing decisions, payroll and payment approval.
| Priority | Opportunity | Judgment |
|---|---|---|
| 1 | Structured weekly reporting | Fix definitions and ownership; test rules first |
| 2 | Request intake assistance | Draft and validate; do not auto-commit scope |
| 3 | Meeting action drafts | Pilot only with approved data and consent |
| Defer | Autonomous client replies | External error cost exceeds the case for speed |
Reporting is recurring, internally visible and easier to compare with a baseline. Intake has more varied language and exceptions. Meeting notes introduce recording and data-handling questions. The priority is based on readiness and risk as well as potential minutes saved.
| Task | Volume | Before / after | Eligible share | Time released |
|---|---|---|---|---|
| Intake | 40 requests | 8 / 5 min | 60% | 72 min |
| Reporting | 10 updates | 18 / 8 min | 80% | 80 min |
| Meeting actions | 12 meetings | 15 / 7 min | 75% | 72 min |
Time released = volume × (baseline handling time − future handling time) × eligible share. Future handling time includes human review. Tasks are defined as separate activities to avoid double-counting. Gross potential is 224 minutes per week. Subtract 45 minutes for ongoing maintenance: net potential is 179 minutes, or about 3.0 hours per week.
At an assumed 48 working weeks, the base case releases 143.2 hours a year. This is available capacity, not a payroll reduction or realised financial return. Implementation effort, software costs and how the capacity would be used must be assessed before approval.
| Scenario | Gross benefit realised | Net min/week | Net hours/year |
|---|---|---|---|
| Conservative | 50% of base estimate | 67 | 53.6 |
| Base assumption | 100% of base estimate | 179 | 143.2 |
The workflow is infrequent, input data is unreliable, review takes as long as the existing task, exceptions dominate, or no owner can maintain it. In the conservative scenario, 45 minutes of maintenance still apply. Actual timings and eligible shares must be observed before relying on either scenario.
Start with weekly reporting only. Use one agreed source table with project ID, owner, reporting period, status, next milestone and blocker. Check missing fields and duplicate project-period records with rules. AI drafting is optional; a template may be sufficient. Any draft must reference its source rows and label missing information.
The operations lead approves every update before publication. Missing owners, conflicting dates and unsupported statements go to an exception queue. No autonomous changes to project commitments or external messages. Keep an editable log and a manual fallback. Use only client-approved data and tools; redact sensitive inputs during testing.
| When | Work | Decision |
|---|---|---|
| Days 1–5 | Observe a full cycle; agree baseline | Is the problem frequent and measurable? |
| Days 6–10 | Clean definitions; prototype one flow | Can ordinary rules do the job? |
| Days 11–20 | Shadow-run alongside manual process | Are outputs complete and supportable? |
| Days 21–30 | Review effort, errors and ownership | Adopt, revise or stop |
Every published statement is supported by source data and approved by its owner. All required fields are present or explicitly flagged. Run at least two reporting cycles and compare handling time including review and corrections. Progress only if net effort falls, no critical factual errors reach publication and a named person accepts maintenance ownership.
A process map, ranked opportunity list, assumptions table, benefit sensitivity and pilot brief. These are sufficient to make the next decision without committing to a large build. This is a fictional analysis demonstration; projected benefits are not client results.
A focused first engagement
Bring one workflow that creates recurring frustration. Leave with a clear map, a prioritised improvement list and an actionable next step.
Five business days from complete inputs.
One consolidated revision included.
Scope: one defined workflow. Software builds, production integrations and broader audits are scoped separately. Benefits are assessed, not guaranteed.
Discuss your workflowAbout Audun
My background combines business analysis at Statkraft, external audit at EY and a master’s degree from the Norwegian School of Economics, specialising in Business Analysis and Performance Management.
At Statkraft, I worked with resource coordination, portfolio reporting and process modelling with business stakeholders. At EY, I worked with financial statements, internal controls and business processes.
Through my own projects, I use Claude and other AI tools to explore ideas, create prototypes and develop practical ways of working. I bring that experience to process assessment and workflow design, with clear boundaries around the technical implementation required.
Based in Norway. Remote collaboration in English and Norwegian.
Estimate the time a workflow could release, including human review and maintenance.
Before we start
Consulting firms and agencies with roughly 10–50 people, several concurrent projects and recurring coordination or reporting problems. Smaller teams can also be a fit if the workflow is clearly defined.
You receive an editable map and prioritised actions. You can implement these internally, agree a separate follow-on scope with me, or use the findings to brief another specialist.
The review determines what is worth doing and what implementation requires. Small prototypes may be appropriate. Production integrations, specialist engineering and ongoing operation require a separate scope and the right delivery capability.
Choose one workflow, identify its owner and bring up to three representative examples. We agree a suitable way to share information before any internal material is provided.
Yes. Delivery is remote in English or Norwegian. Scope, schedule and applicable taxes are confirmed in the written offer.
Start with the problem
Tell me what happens today, the tools involved
and what you want to improve.
Meeting me through Upwork?
Please keep pre-contract communication there.