How can AI help us do this faster?
Add AI to the current process. Optimize one step. Preserve the handoffs.
HUMAN + AI ORGANIZATIONPPT 01 · 01 / 30
LFMA CONFERENCE SESSION A NEW RELATIONSHIP WITH WORKHuman+ AI.
THE ORGANIZATION, REIMAGINED
Leading change and governing
the future of work.
A practical exploration for mission-driven finance leaders.
THE SHIFTPPT 02 · 02 / 30
The way we delegate
is changing.
From people using AI to people managing AI work.
Define the task. Give AI the context. Review the work it brings back.
THE SHIFT IN PROGRESSTHE USEPPT 03 · 03 / 30
A better question.
A different workflow.
Why does this need 14 steps at all?
Redesign around the outcome. Keep people where judgment and relationships matter.
A faster step inside a poor workflow only moves the bottleneck.
THE DIRECTIONPPT 04 · 04 / 30
Inside the job,
the work redistributes.
Processes change first. Roles within those processes follow.
MORE AI PREPARATIONMORE HUMAN DIRECTION
PEOPLE SET THE DIRECTION
People challenge the evidence, weigh consequences, and decide where AI must stop. Judgment and control remain central to the work.
SELECT A PART OF THE WORK TO EXPLOREAI takes more preparation.
People take more direction.
THE CHOICEPPT 05 · 05 / 30
Your role.
Recomposed.
Controlled, ad hoc use is only a beginning. Stronger foundations make greater delegation possible.
YOU DO THE WORK“I create.
“I create.
AI assists.”
You produce the work and make the decisions. AI helps with a specific question or step when you need it.
AN ILLUSTRATIVE FINANCE EXAMPLE
You write the financial commentary and ask AI to help clarify a difficult explanation.
Foundational maturity ↑AI autonomy →
Before delegating more: make the outcome, the limits and human accountability clear.
THE BUSY HUMAN PARADOXPPT 06 · 06 / 30
Busy reacting.
Or ready to act?
Work cultures often reward reactive behavior. AI is often used in the same way.
Fragmented, reactive tasks.
Immediate problems. Isolated knowledge and experience. Siloed information.
Anticipate
Plan · Prepare
Integrate
Plan · Prepare
Integrate
Prepare before urgency.
Integrated processes. Shared knowledge and context. Shared foundations.
THE ORGANIZATIONAL FOUNDATIONPPT 07 · 07 / 30
Make judgment
available to the work.
Shared knowledge helps AI execute consistently without losing the judgment behind the work.
Human judgment
Experience interprets context. People weigh consequences and competing needs. Some decisions need a conversation.
Shared judgment
Record purpose and evidence. Make approval limits and exceptions clear. Agree what a good result looks like.
AI capability
Prepare trusted information and apply documented context. Follow agreed steps within authority. Pause when context or evidence is insufficient.
Human attention stays with exceptions, consequences and accountability.
REALIZING AI VALUEPPT 08 · 08 / 30
Three foundations.
One useful workflow.
Start with one workflow, a clear owner and a measurable outcome.
From searching to reasoning.
Learn to work through decisions with agents, beyond looking up facts. People delegate the right work, describe the goal and context, and discern whether the answer holds up.
Shared context. Shared meaning.
Individual agent capability needs a shared foundation to become enterprise capability. Data models describe how information is stored, while ontologies capture context, semantics, constraints and governance.
Controls that won’t bend.
The agent reasons, but the system enforces what must be exactly right. Fixed controls and safeguards protect sensitive data and critical decisions, so outputs can be trusted every time.
THE THREE FOUNDATIONSPPT 09 · 09 / 30
What agent workflows
stand on.
Agentic workflows ↑
Literacy
DelegateDescribeDiscern
Alignment
Operating modelChange processStandards
Guardrails
Reliable dataPolicyCore systemsAccess & protection
FROM FOUNDATION TO EXECUTIONPPT 10 · 10 / 30
Define. Prove.
Then coordinate.
Defined workflow
A clear outcome. An accountable owner.
Shared foundation
Trusted sources. Clear decision criteria.
Controlled pilot
Review exceptions. Measure results.
Coordinated execution
Connected agent tasks. Human oversight.
Alignment Purpose, workflow design and authority limits
Literacy Training, adoption and continued practice
Guardrails Reliable information, shared definitions and ownership
Evidence and clear accountability support each increase in delegation.
A REAL USE CASEPPT 11 · 11 / 30
From buried contracts
to frontline knowledge.
LCS NORTHWEST / PROGRAM MANAGEMENT
01Understand the contents of programs, their requirements and who is eligible.
02Bring unstructured information buried in contracts to frontline teams.
03Improve reliability, speed of access and consistency of information.
THE LCS NORTHWEST JOURNEYPPT 12 · 12 / 30
“We have all
the data we need.”
“All budgets for programs are in SharePoint.”
“I have all the contracts for my district in this folder.”
“Our Contract Management system has all the information about our contracts.”
THE LCS NORTHWEST JOURNEYPPT 13 · 13 / 30
Then the cracks
begin to surface.
OWNERSHIP“Who is Maria’s supervisor?”
SHARED DEFINITIONS“I don’t work this program, I work the ORCA Program.”
CURRENT INFORMATION“This contract is expired.”
ACCOUNTABILITY“Why is our CEO the owner for all these contracts?”
THE LCS NORTHWEST JOURNEYPPT 15 · 14 / 30
The workflow.
In motion.
THE LCS NORTHWEST JOURNEYPPT 14 · 15 / 30
The same foundations.
Made operational.
Agentic workflows
Literacy
DelegateDescribeDiscern
Alignment
Operating modelChange processStandards
Guardrails
Reliable data
Policy
Operational systems
Access & protection
Reliable data
Policy
Operational systems
Access & protection
IN A LIVE SITUATIONPPT 17 · 16 / 30
When the need
is immediate.
A CASEWORKER’S EXPERIENCE
“On Thursday I had a situation with one of my clients, who is a single mom with an 11-month-old child who called me and said that she is in a park with her baby, and nowhere to go.
… I asked the AI Agent what to do in a situation like this, and I was pleasantly surprised by how helpful it was.
It guided me through clear immediate steps, including phone numbers to call and organizations to contact (without direct link) such as mental health crisis lines, housing resources, local shelters, drop-in centers, and Basic Needs Assistance programs.”
MISSION CAPACITYPPT 18 · 17 / 30
03 / MISSION CAPACITYWhat is
What is
human attention
worth?
Every hour spent moving information is an hour not spent applying judgment, building relationships, or advancing the mission.
IMAGINE THE CAPACITY
hours / year
to put back into the mission.
Illustrative scenario, not a forecast.Hours × people × 46 working weeks.
THE SHIFT, REVISITEDPPT 19 · 18 / 30
Stronger foundations.
A different role.
As foundations strengthen and more judgment can be shared with AI, roles and organizational structures can move toward directing and orchestrating.
Author
I do the work.
AI assists when needed.
Editor
AI drafts.
I improve and approve.
Director
I define the outcome.
AI completes the task.
Orchestrator
I design how people
and agents work together.
Foundational maturity ↑AI autonomy →
THE HUMAN SIDEPPT 20 · 19 / 30
Change the work.
Bring the people.
Fear deserves a conversation—and a practical role in shaping what comes next.
01“Will I become irrelevant?”+
Shift toward being an AI-enabled operator: direct people, systems, and agents to deliver outcomes. Build AI fluency alongside domain knowledge. AI can handle repetitive drafting, research, analysis, reporting, and documentation, while you own the workflow, judgment, and outcome.
02“Can I challenge a decision?”+
Build AI fluency alongside domain knowledge. Bring your experience to governing AI and interpreting its decisions. Keep the work transparent, name the human owner, and preserve the evidence needed to challenge a result.
03“Will work just get more intense?”+
Use AI to remove low-value activity and unnecessary handoffs. Agree how recovered capacity will support meaningful work, and measure the effect on the team.
04“Can I trust the intention?”+
Become an active architect of the change. Leadership must explain human and AI accountabilities, policies, and expectations. People should help redesign the processes they know, govern outcomes, and oversee AI use in their own domains.
PREPARING FOR THE FUTUREPPT 21 · 20 / 30
A team of agents.
A human at the centre.
The human workflow owner sets the outcome, standards, authority and stop points.
✳Human ownerDirection & accountability
ROLE 01 / 05
Intake agent
Collects requests and clarifies scope so the right work starts with the right context.
ILLUSTRATIVE ROLE BOUNDARIES
PERMITTED WORK
Organize requests, identify missing information, and prepare a clear brief.
HUMAN STOP POINT
A person resolves conflicting priorities and approves material changes in scope.
Manage an outcome through a designed operating model: named agents, defined work, clear controls and human accountability.
INDUSTRY DIRECTIONPPT 23 · 21 / 30
The shift is
already happening.
30min
Legacy refactoring,
previously two weeks.
Engineering time compressed, with teams still responsible for quality. Product teams also report turning weeks of research into hours.
15+days
Manual effort removed
per financial close cycle.
AI supports the preparation. Finance managers retain final posting approval.
16million
AskHR user messages
during 2025.
Multiagent orchestration supports more complex HR transactions behind one interface.
<30hours
Time saved through
coordinated operations.
An agent team supporting time entry and approval, invoicing, reconciliation and payroll.
REPORTED EXAMPLES · RESULTS VARY BY WORKFLOW AND ORGANIZATION
A REAL-WORLD AGENTIC TEAMPPT 24 · 22 / 30
Critical analytics.
Constrained headcount.
THE TRADITIONAL TEAM
Project & triageDomain expertiseData expertiseGovernanceEngineeringReportingQuality assuranceSecurity
Different skills. Multiple employees. Manual coordination.
Limited headcount budget. Variable demand, fixed costs. “Do it all” roles become single points of failure.
Lead
agent
agent
Human leader
Domain
agentsDomain
agentsGovernanceComplianceQuality
assuranceSecurity
agentsDomain
agentsGovernanceComplianceQuality
assuranceSecurity
ReportingDataInsightsEngineering
One human leader. Multiple systems. Hundreds of users.
Agents have defined responsibilities. Outputs pass through human decision gates.
POLICY · PROCESS · PEOPLE · GOVERNANCE · TECHNOLOGYPPT 26 · 23 / 30
The whole organization
changes with the work.
| Foundation | Policy | Process | People | Governance | Technology |
|---|---|---|---|---|---|
| Literacy | Define the process | AI and human skills; optimize jobs; aligned training and standard work; onboarding. | |||
| Alignment | Redesign workflows to use AI. | Share knowledge and experience; remove siloed decisions; build common understanding. | Ontology: definitions and dictionaries; models; authorization flows. | Data contracts, models, flow and integration. | |
| Guardrails | Acceptable use; privacy and breach handling. | Enforce policy in redesigned processes; audit and monitor AI use. | Monitor human usage; authorize human decisions. | Monitoring and auditing; role-based authority; least privilege. | Tools and platforms; licensing and budgets; authentication and authorization. |
What changes would you expect in your organization?
PEOPLE / HOW ROLES EVOLVEPPT 27 · 24 / 30
Knowledge and control:
still human.
Execution may shift to the machine. Human roles expand in accountability and responsibility.
Own the result.
- Measurable service quality and speed
- Evidence-based controls and governance
- Domain expertise and insight
- Approval of actions and agent recommendations
Improve the system.
- Define and share knowledge and experience
- Optimize business workflows
- Refine and coach agent skills
- Audit and monitor
CONTROLLED DELEGATIONPPT 22 · 25 / 30
Treat Agents like employees.
Roles need boundaries.
→Reducing risk exposure←
Job
A defined outcome
to own.
Identity
A named account
and permissions.
←
Manager
A human to review
and approve.
←
Record
An auditable history
of what happened.
Can do
Prepare work, summarize evidence
and recommend next steps.
Must escalate
Policy exceptions, material risk
or low-confidence outputs.
Cannot
Self-approve, exceed authority
or bypass audit and controls.
Accountability stays human.
WHERE NEXTPPT 29 · 26 / 30
What does
standing still cost?
01Will your organization be expected to become more efficient as AI advances?
02Will boards, regulators, donors and others demand more stringent measures and results?
What are your thoughts?
YOUR NEXT MOVEPPT 30 · 27 / 30
Start with
one piece of work.
01 What work draws people away from your mission?
02 What truly needs human judgment or relationships?
03 What could you redesign if capacity were no constraint?
Design the future workflow, rather than copying today’s steps. Make the outcome measurable.
A CONVERSATION ABOUT WHAT COMES NEXTPPT 31 · 28 / 30
Questions?
TAKING THE NEXT STEPPPT 32 · 29 / 30
Turn possibility
into priorities.
Where should AI
make its next impact?
Turn stakeholder interviews, business objectives, operational data and opinions into data-backed priorities that drive ROI.
Spark uses AI to identify valuable, achievable use cases before time and budget are spent.
LET’S WORK TOGETHERPPT 33 · 30 / 30
More human possibility.
Thank you.
hello@proactiviti.com ↗HQ / SEATTLE
1400 NW Leary Way, Suite 400
Seattle, WA 98107
+1 206.800.8390
Proactiviti LLC
