AIM · AI Interim Management

Your AI function.
Run by us, until you can.

AIM is interim management for the AI function. We take responsibility for AI in your operation for a defined period — we map the processes that currently live in files and in people's heads, put them into a system, and run them until your team runs them without us. Diagnostic, plan and implementation are contracted separately, and each one stands on its own.

ModelInterim management
Phases3, separable
Delivered onEOS
FitMid-market and up
EOS · Live
Before
Spreadsheets everywhere.
Scattered SOPs · manual process
Phase 01
Diagnostic.
Process audit · AI opportunities
Phase 03
EOS installed.
E
Agents · workflows · apps
After
Operation running on its own.
Runs today1,842
Hours saved68
Accuracy94%
Dashboards · alerts · reporting

The operation lives in what the team knows.

Most AI programs stall for a concrete reason: the processes they are meant to improve are not written down anywhere. They exist across documents, spreadsheets and the experience of the people running them. That is hard ground to build a system on. AIM structures the operation first. AI comes in after that, and it fits.

01

Procedures split between files and tacit knowledge.

The way work actually gets done is distributed across documents, spreadsheets and the memory of the people who do it. Every departure, promotion or new hire turns knowledge transfer into an expensive draw on management time.

02

Follow-up spread across channels.

Open items circulate through chat, email and standing meetings. Open cases advance with no structured tracking. The evidence that something was executed is scattered, and reconstructing it after the fact costs real hours.

03

Misses that surface late.

Steps that were skipped or done wrong tend to appear in an audit, a customer complaint or a month-end close — by which point the cost is already incurred. That exposure is almost never quantified.

What AIM proposes
The operation stops depending on individual knowledge.
It starts running on the system.

Three phases, separable.

Each phase is contracted on its own. You can start with the diagnostic, move to the plan when the findings justify it, and implement after that. There is no commitment to all three up front. Each phase produces a deliverable that stands alone.

Phase 01
Diagnostic.
3–4 weeks · paid

We document the SOPs you have today, map the processes that matter, identify where AI actually pays and where it does not, and audit the data you can reach. This is not a generic assessment. It is your operation, examined.

  • Interviews with function leads
  • Process and SOP mapping
  • Audit of existing data sources
  • Prioritized matrix of AI opportunities
  • Executive report · 20–30 pages
Phase 02
Plan.
2–3 weeks · paid

We turn the diagnostic into a roadmap you can execute: what gets built first, what comes after, on what stack, budgeted by milestone, with the internal resources required and the metrics that define success. A plan that holds up in front of a board.

  • 90 / 180 / 365-day roadmap
  • Proposed technical architecture
  • Implementation budget by milestone
  • Success metrics per initiative
  • Risk analysis and mitigations
Phase 03
Implementation.
3–12 months · paid

We build. Conversational agents, automated workflows, internal apps with AI built in, LLM analysis across your data, executive dashboards. Each month you get a report on what ran, what it saved and what comes next.

  • Execution in monthly sprints
  • Progressive handoff to your team
  • Training for the internal team
  • Governance and usage policy
  • Monthly executive report
Separable, not bundled

Each phase is quoted and signed separately. You can run only the diagnostic and take it to another firm to build. You can combine diagnostic and plan and implement with your own team. The point is that the decision gets made on evidence, instead of committed to in advance.

The system we deliver on.

AIM runs on EOS — our operating system for running a company on evidence. It is not an abstraction. It is a concrete equation: your SOPs go in, the calendar decides when each one is due, and three lanes (human, AI, engine) execute them, leaving an auditable trail.

The EOS equation
CALENDAR × SOPs + HUMAN   +   AI   +   ENGINE Auditable operation

In EOS every SOP is a contract with five fields: who owns it, what goes in, what it does, what comes out, and how you know it was done right. The calendar decides when it is due. The three lanes decide who executes it.

Every step in the right lane.

A single process mixes three lanes of execution: the system handles what repeats, AI prepares the draft, and your team decides what requires judgment. No process starts in the automatic lane. It earns its way there once the rule is stable and the corrections have settled into one.

Lane 01
Manual.

A guided checklist with evidence and approvals. Measurable and auditable, even though a person still does the work.

Physical count · credit approval
Lane 02
AI-assisted.

AI prepares the draft; your team approves, corrects or discards it. Nothing moves forward without a human sign-off.

Weekly report · document extraction
Lane 03
Automatic.

Executed in code. The same every day, without variation. The person moves from doing the work to reviewing the result.

Data reconciliation · weekly OTB
01
Capture
We capture the SOP.

We pull the process out of wherever it currently lives — documents, spreadsheets, the person who has always done it. It gets structured into five fields: who, what goes in, what it does, what comes out, how it is validated.

Manual
02
Install
It goes into EOS.

The SOP goes in as an executable contract. It gets an owner, a second who can cover it, the evidence it is expected to produce, and a validation rule. The process is auditable from day one.

Manual
03
Cadence
You declare the cadence.

EOS schedules it by its own rule — a fixed date, a triggering event, a condition being met. The calendar stops depending on someone remembering.

Manual
04
Collaboration
Human and AI work together.

AI proposes drafts, extractions and classifications. The person decides: approve, correct or discard. Every interaction is recorded.

AI-assisted
05
Learning
EOS distills the rules.

Every human correction becomes a reviewable rule. In month one you correct twenty details; by month six, two or three. The process improves without being reprogrammed.

AI-assisted
06
Autonomy
EOS executes, the human supervises.

Once the rule is stable, the process moves to the automatic lane. The person stops operating and starts supervising: reviewing output, approving exceptions, validating changes.

Automatic
Not every process reaches step 06 — and not every process should. Some settle at structured manual, others at stable AI-assisted. What matters is that the decision to move a process forward is made on the data, not on faith.

Nine service lines, one team.

This is not "AI consulting." It is a team with nine service lines, deployed according to what the diagnostic finds in your operation. Everything produced for your business is yours — the code, the prompts, the documentation, the reports.

01

Process audit and SOPs.

We document how the company runs today, where the friction sits, and what should be automated with AI versus what should simply be a better process.

02

Conversational agents.

For customer service, B2B sales, internal support, lead qualification. With guardrails, your brand voice, and handoff to a person when the conversation calls for it.

03

Automated workflows.

n8n, Make, Zapier, or custom in Python or Node. Quoting, order follow-up, inventory alerts, recurring reports.

04

Internal apps with AI built in.

Executive dashboards, analyzers, spec-sheet generators, internal tools. Built around your operation.

05

LLM analysis across your data.

Across your ERP, CRM, ecommerce, messaging and email. Extraction, classification, and an automatic weekly executive summary.

06

Prompt library.

Tested prompts organized by use case, versioned, with performance metrics. Reusable by your entire internal team.

07

Training for your team.

Working sessions, not lectures. Your people learn to use and modify what we build. By the end, the internal team runs AIM and EOS.

08

Governance and policy.

Explicit rules for responsible use, data privacy, the limits of what AI is allowed to decide, and exception protocols. Auditable.

09

Monthly executive report.

What ran, how many hours it saved, what failed and why, what comes next month. Four to six pages, ready for the board.

EOS in production.

Three views of EOS running inside a company. AIM installs it in phase 3. Switch tabs to see the operating dashboard, an agent handling a live customer, and a workflow laid out end to end.

eos.your-company.com · overview
E
EOS
Operations
Overview
Automated OTB
Demand Forecast
Routing
Pricing
System
Agents
Workflows
Reports
All running Last run · 2 min ago

Overview · EOS

Live · 12 active processes
Runs today
1,842
↑ 12% vs yesterday
Hours saved
68 hrs
this month
AI accuracy
94.2%
last 30 days
Failures
2
need review

Latest runs

last 2 hours
Weekly OTB recalculatedCategories: 12 · SKUs: 3,240
2 min ago
-4.2 hrs
Delivery routes optimized28 routes · 340 stops
18 min ago
-1.8 hrs
!
Dynamic pricing · line B+3.2% adjustment proposed · human review
34 min ago
pending
Sales call analysis142 calls · 3 service flags
1 hr ago
-3.5 hrs
AR reminders sent84 accounts · 12 replies
1.5 hrs ago
-2.1 hrs

Active conversations

Dave R. — Midwest SupplyOrder status
Brookline GrocersReorder quote
Karen DoyleAddress change
Alton Trading Co.Credit status

Dave R. · Midwest Supply Co.

Agent active
Morning — need to know when my 200 units of Cascade Roast land.
EOS · Sales Agent Morning, Dave. Found your order #48720 from June 18. It ships out of the main DC today at 12:30 PM. Estimated delivery tomorrow between 9 and 11 AM at your Riverside location. Sending you the tracking number and the driver's phone number now.
Good. While I have you — what would 50 more of the same run me?
EOS · Sales Agent At your current volume discount, 50 additional units come to $1,120.00, net 30. I can add them to the same shipment if you confirm before 11:45 AM. Want me to go ahead?
Dave is typing

Automated OTB · workflow

v2.4 · ran 2 min ago
Input
ERP · 90d sales
SAP HANA
Input
Current inventory
Odoo
Input
Base forecast
Prophet
LLM · OTB engine
Synthesis and decision
Claude Sonnet
Output
Purchase budget
by category
Alert
Human exceptions
Slack · Buyer
01 · Ops
The EOS operating dashboard. Execution KPIs, recent runs with hours saved per case, and alerts for the exceptions that need a person to look at them. One place where management sees what is running and what failed. Updates live.
Illustrative interface. The numbers shown are representative, not pulled from a live client.

Where AIM has the most impact.

These are the processes where AI automation has the best impact-to-effort ratio in a commercial operation. The figures below are modeled estimates built from a 114-process catalog at a single distributor. They are not results we have delivered to a named client, and they are here to size an opportunity. The range that applies to your operation is established in the diagnostic.

OTB · Purchasing

Automated Open To Buy.

Open To Buy, the weekly purchase budget. EOS combines sales history, current inventory, forecast and seasonality to recalculate it by category and SKU. It alerts the buyer when a call needs to be made. Exceptions are documented.

-70% Buyer time spent
on manual OTB
Forecast · Demand

Demand prediction by SKU.

Prophet plus an LLM over 90 to 365 days of sales and outside factors — calendar, weather, campaigns. More accurate than a manual spreadsheet, and explainable at the SKU level.

+15% Accuracy vs
manual forecast
Logistics · Routing

Daily route optimization.

Every morning EOS recalculates distribution routes against confirmed orders, vehicle capacity, delivery windows and expected traffic. The driver gets a finished route.

-20% Miles driven
per day
Pricing

Dynamic pricing with guardrails.

Reads public competitor pricing, historical elasticity and margin targets. Proposes per-SKU adjustments with a stated rationale. Never changes a price without human approval.

+4-8% Average margin
on selected SKUs
Sales · QA

Sales conversation analysis.

EOS processes rep calls and chats. It flags unresolved objections, prices communicated wrong, and opportunities left on the table. Sales management gets the case and the evidence.

3.5x Volume reviewable
vs manual QA
AR · Collections

Smarter collection reminders.

Reads each account's payment pattern and adjusts the message, channel and timing of the reminder. Text for some, email for others, a human call for the accounts that warrant one.

-12 days Average DSO
vs baseline
Product · Content

Spec sheets at volume.

From raw supplier catalogs to finished product pages with SEO, keywords and comparisons, in your brand voice. A hundred SKUs processed in hours instead of weeks.

50x SKUs processed
per day vs manual
Management · Reporting

Automatic weekly executive report.

EOS joins sales, margin, inventory, receivables and logistics into a three-page narrative report ready for the management meeting. With findings, not just tables.

-6 hrs Of analyst time
per week
Modeled from a 114-process catalog, not measured across a client portfolio. Actual impact depends on the state of your data, the maturity of the process and your internal team's capacity to execute. Every diagnostic establishes the baseline and the expected range before any scope is committed.
Selected clients
Elex, S.A. Greenergyze Habitare Psicometrikas Cash FP

Not everything gets automated. That's fine.

We ran EOS against a 114-process catalog from a single full-line distributor: finance, tax, payroll, inventory, warehouse, collections, sales. This is how they split across the three lanes. It is a starting estimate, not a commitment.

~20of 114
Automatic within months
The part that sells itself.

Processes that consume hours of expensive people today and can run on their own with evidence. None of them need AI — they need the ERP to be readable.

~55of 114
Naturally AI-assisted
The heart of it.

Reconciliations with classified variances, reports with alert bands. None of these automate in one step; all of them improve month over month as history accumulates.

~39of 114
Stays human
The part nobody advertises.

Meetings, calibration, credit decisions, negotiation, physical receiving. A third of the operation can't be automated, and shouldn't be. AIM organizes these; it does not replace them.

The read

A product that only handles what fully automates handles roughly one process in six. AIM governs all three lanes: the whole system, including the parts that stay human.

From kickoff to the first measurable return.

This is the standard rhythm when all three phases run together. If you contract only the diagnostic or only the plan, the schedule adjusts but the rigor does not.

Week 1 · Kickoff

Alignment and setup.

Session with leadership. We agree on the three to five processes that go into the diagnostic. Access setup across ERP, CRM, messaging and ecommerce. NDA signed. Opening session with the operating teams.

Weeks 2–4 · Diagnostic

SOPs captured, data audited.

One-on-one interviews with function leads. We map the flows as they actually run. We audit the data sources. We separate quick wins from structural friction. You get a 20- to 30-page executive report.

Weeks 5–6 · Plan

Prioritized and architected.

Findings become an executable roadmap. Prioritized by impact against effort. Proposed technical architecture. Budget by milestone. Success metrics. Risk analysis. Board presentation if the decision calls for one.

Weeks 7–12 · Sprint 1

First process live in production.

We pick the highest impact for the lowest risk. Solution design, prototype, iteration, deployment. By the end of the sprint something is running inside your operation with before-and-after numbers attached.

End of month 3

First monthly executive report.

What was built. What ran. What failed and why. How much time it saved. What comes next month. Session with the board or management committee. Decision on whether Sprint 2 proceeds.

Where it fits.

AIM does not fit every company. We would rather say so before a sales conversation than during one. The first list is when an implementation has a high probability of working. The second is when something else would serve you better.

✓ It fits when...

The right profile.

  • Mid-market or larger, with at least 30 people on payroll
  • The operation already runs on an ERP or CRM, even partially
  • There are at least 12 months of accessible historical data
  • You want a measurable result in 90 days, not a multi-year horizon
  • Leadership expects to change how the operation works, not only what software it runs on
  • The company can commit 3–8 hours a week of internal team time during phase 3
✗ It doesn't fit when...

Another route serves you better.

  • You want one specific AI build — a single agent, with no system around it
  • The decision comes down to price alone
  • There is no willingness to change internal process, only to automate the current state
  • The data is fragmented and there is no appetite to consolidate it
  • Results are expected without metrics or handoff to the internal team
  • Leadership isn't willing to let a model touch any decision

What it is not. Equally important.

Tools that promise everything usually deliver the easy third. AIM does one thing: it makes your operation run. These are the four limits we hold to, and the ones other firms tend to sell past.

×

Not another task manager.

Nothing on the list is made up. Every item comes from a process you own, with an owner and expected evidence. Ad-hoc work stays in your chat, where it belongs.

×

Not a replacement for your ERP or your books.

SAP, Odoo or whatever you run stays the system of record. EOS organizes the human operation around it. It reads the ERP through a local replica — it never writes to it.

×

Not a brittle bot clicking through screens.

We do not automate someone else's user interface. We hand your people the data already prepared: less brittle, and auditable.

×

Not AI making decisions on its own.

The AI proposes; your people validate. Anything that must come out identical every time is done in code, not by a model.

The rule

AIM has a defined and bounded scope: making the operation run, with evidence, with an owner, and on a calendar. It doesn't go beyond that.

KOJ Consulting installs it.

A boutique firm. A systems practice.

KOJ Consulting works in analytics, business intelligence, applied AI and systems design. We work with mid-market and larger companies, and engagements run remotely. Our working hours overlap the full U.S. business day, across all four time zones.

AIM is our interim management practice for the AI function: how we get operating AI into companies that need a result in months, without the company having to hire and sustain an internal AI team first. It is delivered on EOS, our proprietary system.

◼ Related prior work

  • Psicometrikas — AI-assisted assessment platform
  • Habitare — identity systems and architecture
  • Cash FP — financial analytics for small businesses
  • Greenergyze — B2B operating systems
  • Elex — process automation, 120+ SOPs in production
  • HubSpot, Odoo and AWS implementations across 20+ clients

What people always ask.

What does AIM cost? +
We don't publish prices, because the number moves with the shape of the operation. Pricing scales with two things: how many processes come into scope, and how many people touch them. After a call and a short scoping exercise, you get a quote within five business days. The diagnostic is quoted and signed on its own — if you finish it and decide not to continue into plan or implementation, nothing further is owed and there is nothing to unwind.
Can I contract only the diagnostic? +
Yes. The three phases are separable. You can run the diagnostic with us and take it to another firm to build. You can run diagnostic and plan and implement with your own team. Or you can run all three with KOJ. Each phase produces a deliverable that stands on its own.
How long until we see results? +
Diagnostic: three to four weeks. Plan: two to three weeks after that. First automation live in production: somewhere between week 7 and week 12 from kickoff, when all three phases are contracted. The first executive report with measurable hours saved arrives at the end of month three. Quick wins can land earlier if the diagnostic turns up something simple.
What happens when the engagement ends? +
Everything built for your business is yours: code, prompts, workflows, documentation, dashboards, models and playbooks. EOS itself — the platform the work runs on — remains KOJ property and continues under license. During phase 3 we run a progressive handoff so your internal team can modify and maintain what is installed. If you choose not to renew, the operation keeps running without depending on us. You can bring us back for specific blocks of work.
Do we need clean data to start? +
No. The data needs to be reachable, not perfect. The diagnostic audits what exists, what is missing and what can be recovered. If fragmentation is significant, the plan will likely include a consolidation stage before any agents get deployed. The gating question isn't data readiness. It's whether the company will commit to consolidating what's fragmented.
How does a remote engagement actually run? +
Fully remote, and built that way rather than adapted to it. Kickoff and the sessions that matter run over Zoom or Meet; day-to-day collaboration happens in Slack and email. Our working hours overlap the entire U.S. business day across all four time zones. Questions get answered the same day rather than the next one, which is the difference that actually shows up in a schedule. On-site sessions can be arranged for the phases where being in the room changes the outcome.
What about the security of our data? +
We sign an NDA before we touch any data. The providers we use are enterprise tier — Anthropic, OpenAI, AWS — under contracts that prohibit using customer data for model training. When sensitivity requires it, models are deployed inside your own infrastructure on AWS or Azure. Governance is documented in the plan and audited monthly.
Are you going to replace our people? +
No. AIM makes the team you have go further; it does not replace them. What gets automated is repetitive and low-judgment: reports, reminders, standard quotes, OTB. The team gets freed up for higher-value work: complex decisions, key accounts, negotiation. AIM is not a headcount reduction play. The cases where a client reduced staff after implementation are the exception, not the pattern.
Next step · Diagnostic

Put AI where the operation loses time.

Six short questions to tell whether AIM is a fit. If it does, we send scoping and a quote for the diagnostic within five business days.

Question 1 of 6
Received

Got it. We're on it.

Your information is in. You'll hear back by email within one business day with next steps. The diagnostic doesn't commit you to anything past it.

Esta página también está disponible en español.

Ver en español