These are prototypes, and the intelligence is emulated. No call to Jev or TypeSafe AI produced any number on this site. Everything runs against an offline stand-in that speaks Jev's published contract (POST /v1/systemone, the Noul / Choice / Score primitives) and returns probability distributions calibrated by construction against hidden labels in a synthetic corpus. Every company, deal, filing and hand below is invented. Swapping the emulator for the real model is one flag: --backend live. Not affiliated with, or endorsed by, TypeSafe AI.
Prototype 2 · Spin-out replatform
24 systems, 18 months, no IT department
A $640M brand line spinning out of Continental Brands, a multinational CPG. Close is 1 April 2027. Brightline has 1,400 employees, three plants, two distribution centres, and no IT organisation of its own. The Transition Services Agreement keeps the lights on for
18 months after close. From month 19 the fee doubles to
$2.9M a month and the parent can terminate on 90 days notice.
Nobody has the dependency graph; it is implicit in two dozen arrangement memos.
Systems
24
$33.8M/yr to run
TSA window
18 mo
then $2.9M/month
IT budget
$38.0M
to stand it all up
Judgments
168
in 24 requests, $0.0010
The mechanism
The plan is a dependency graph nobody has written down
Every question that decides this programme — is it day-one critical, what blocks it,
can the contract novate, how entangled is the data — is a judgment somebody makes by
reading a memo. Twenty-four systems, seven questions each: high-volume, repeated, typed
decisions over a shared state, which is exactly the shape a System One model is for.
Nobody at Brightline holds the dependency graph. It is implicit in two dozen memos written by different people, and it is the only thing that decides the plan.
The plan nobody wrote down. Each arrow is one blocking_dependency answer; the waves are what falls out when you sort 24 of them topologically. Highlighted systems are the five latest finishers — the critical path.
One system, close up
The ERP decides the date
The ERP is the long pole, and its disposition is not a preference — it is a reading of whether anything can transfer. Continental will not licence their SAP client, so Jev puts 66% of the mass on rebuild. The gap between the branches is under three months on the ERP itself, and every one of those months moves the 4 systems queued behind it — which is how a single disposition call ends up deciding the programme date. Each trial draws from this distribution rather than assuming the modal answer, which is why the finish date has a spread at all.
The decision
Five ways to sequence it
None of these levers is a model output. The model says what each system is;
the strategy says what to do about it. Narrow the scope, buy instead of build, switch
off what nobody uses, or fund enough parallel teams to start ahead of a dependency.
Each run is 20,000 trials drawing from the judgment distributions.
When the last in-scope system leaves the parent
Median marked; bar spans the 10th to 90th percentile.
best oddsalternativedashed rule — TSA expiry
Probability of clearing the TSA in time
Total cost: TSA fees, penalties, and what the strategy buys
What each strategy actually does
Everything, in order24 systems in scope · P50 20.6mo · 25% on time · $35.4M
Move all 24 systems, respect every dependency, accept the date that falls out. The plan a programme writes before anyone has counted the months.
Day-one first15 systems in scope · P50 19.3mo · 38% on time · $32.2M
Only what the business cannot open without has to clear the TSA. Everything else moves to a narrowed extension and gets done in year two.
Buy over build24 systems in scope · P50 20.5mo · 27% on time · $39.2M
Wherever the vendor will sell us a standalone tenant, take it instead of rebuilding. Costs licence money, buys months.
Narrow and parallel best odds15 systems in scope · P50 17.2mo · 58% on time · $35.9M
Only what the business cannot open without, bought rather than built where a vendor path exists, anything unused switched off, and enough parallel teams to work ahead of the dependencies. Every lever at once.
Buy parallelism24 systems in scope · P50 17.8mo · 52% on time · $37.8M
Same scope as the naive plan, but fund enough parallel teams and systems integrators to start each workstream earlier against its prerequisite.
The finding
The honest answer is that 18 months is the wrong number
The naive plan — move everything, respect every dependency — clears the window
in 25% of trials. Pulling every lever at once gets that
to 58%, and even then the 90th percentile lands
5 months past expiry at $2.9M a month. Buying instead of
building barely moves it, because the critical path runs through master data and the ERP,
and those are precisely the two things Continental will not licence.
That is worth knowing before the TSA is signed rather than in month 14. The
decision this analysis actually supports is not which systems to move first — it is
to negotiate a longer window, or a narrower service list, while there is still negotiating
leverage. A point-estimate plan cannot tell you that; it produces one date and no sense of
whether the date is achievable.
The payoff
Which judgments the date is most sensitive to
A system moves the finish date in proportion to how uncertain its own duration is and how
much of the programme sits behind it. That product is where an architect's week is worth
spending — not on the systems with the biggest run cost.
The same confidence gate as every other prototype, with the thresholds set by what the judgment moves rather than by a global default.
System
Run cost
Day-one p
Conf
Schedule variance
Route
Data platform & warehouse
$2.4M
0.17
0.71
16.5%
human
ERP core (finance, order-to-cash)
$6.2M
0.83
0.59
12.1%
human
General ledger & consolidation
$1.4M
0.89
0.38
9.0%
human
Warehouse management
$2.8M
0.84
0.70
8.5%
human
Core master data management
$1.9M
0.84
0.49
7.6%
human
Identity & directory
$900k
0.90
0.61
7.2%
human
Quality management (FDA scope)
$1.1M
0.78
0.63
5.7%
human
BI & management reporting
$700k
0.11
0.53
4.6%
human
Inventory & demand planning
$1.1M
0.15
0.52
4.1%
human
Lab information management
$600k
0.15
0.58
4.1%
human
The check
Does the confidence mean anything?
Stated probability vs observed accuracy
Scorecard
Judgments scored
168
Accuracy
81.5%
Brier score
0.139
Expected calibration error
0.032
Accuracy when it claims ≥ 0.8
88.5%
Accuracy when it claims < 0.6
57.1%
The evidence
All 24 systems, and what was read to judge them
System disposition
24 systems, seven typed judgments each, in 24 requests.
Identity & directoryRun cost $900k human needed
day_one_critical0.90
tsa_exit_risk3.84 / 4 · 0.72
migration_complexity2.90 / 4 · 0.61
approachsaas_replace · 0.73
data_gravity1.02 / 4 · 0.75
blocking_dependencynone · 0.84
vendor_transferable0.03
Every Brightline employee authenticates against Continental's corporate tenant. SSO integrations, group membership, device trust and conditional-access policy all live there. Continental's security team has said plainly they will not extend directory access...
Network & WANRun cost $2.1M human needed
day_one_critical0.85
tsa_exit_risk2.98 / 4 · 0.40
migration_complexity2.94 / 4 · 0.62
approachlift_and_shift · 0.66
data_gravity0.12 / 4 · 0.78
blocking_dependencynone · 0.53
vendor_transferable0.18
Plants and DCs sit on Continental's MPLS backbone with circuits contracted centrally; site firewalls are managed from their NOC. The carrier confirmed circuits can be novated but quoted 16 weeks per site, and two of the three plants have only one physical c...
Device provisioning & endpointRun cost $1.6M human needed
day_one_critical0.86
tsa_exit_risk2.97 / 4 · 0.89
migration_complexity2.11 / 4 · 0.61
approachsaas_replace · 0.49
data_gravity1.17 / 4 · 0.43
blocking_dependencyidentity · 0.47
vendor_transferable0.12
Laptops are imaged and managed by Continental's MDM, tied to their identity tenant. Around 1,100 machines are Continental-owned assets transferring with the deal; the other 300 are leased under a Continental master lease with no assignment right.
Site Wi-Fi & badge accessRun cost $400k review
day_one_critical0.98
tsa_exit_risk1.93 / 4 · 0.71
migration_complexity1.01 / 4 · 0.56
approachsaas_replace · 0.74
data_gravity0.17 / 4 · 0.71
blocking_dependencyidentity · 0.60
vendor_transferable0.97
Access points at all five sites are Continental-owned hardware on their controller, authenticating against their RADIUS which points at their directory. Badge readers use the same identity source. Hardware is fully depreciated and Continental has offered to...
ERP core (finance, order-to-cash)Run cost $6.2M human needed
day_one_critical0.83
tsa_exit_risk3.83 / 4 · 0.70
migration_complexity3.71 / 4 · 0.59
approachrebuild · 0.38
data_gravity2.76 / 4 · 0.53
blocking_dependencymdm · 0.80
vendor_transferable0.12
Brightline runs inside Continental's global SAP instance, sharing a client with four other brand lines. Chart of accounts, tax configuration and intercompany postings are all shared. Continental's finance team estimates the carve-out of Brightline's master ...
Warehouse managementRun cost $2.8M human needed
day_one_critical0.84
tsa_exit_risk2.65 / 4 · 0.38
migration_complexity3.82 / 4 · 0.70
approachdefer · 0.60
data_gravity1.86 / 4 · 0.12
blocking_dependencyerp · 0.37
vendor_transferable0.75
Two DCs run Continental's WMS, tightly coupled to the shared ERP for inventory postings. The Ohio DC also handles volume for two Continental brands that are staying, so the physical facility is being split as well as the software.
General ledger & consolidationRun cost $1.4M human needed
day_one_critical0.89
tsa_exit_risk3.86 / 4 · 0.77
migration_complexity3.40 / 4 · 0.38
approachsaas_replace · 0.56
data_gravity2.98 / 4 · 0.55
blocking_dependencyerp · 0.79
vendor_transferable0.09
Consolidation runs in Continental's group reporting tool. Brightline has never closed its own books as a standalone entity and has no audited opening balance sheet yet. External auditors need a clean first close within one quarter of separation.
Inventory & demand planningRun cost $1.1M human needed
day_one_critical0.15
tsa_exit_risk2.84 / 4 · 0.61
migration_complexity3.01 / 4 · 0.52
approachsaas_replace · 0.59
data_gravity2.62 / 4 · 0.35
blocking_dependencydata_platform · 0.44
vendor_transferable0.20
Planning runs on a Continental-built tool on top of their data platform. Eighteen years of demand history lives there and the forecasting models are tuned on it. Continental will provide a data extract but will not licence the tool.
Core master data managementRun cost $1.9M human needed
day_one_critical0.84
tsa_exit_risk3.88 / 4 · 0.77
migration_complexity3.50 / 4 · 0.49
approachsaas_replace · 0.55
data_gravity3.71 / 4 · 0.58
blocking_dependencynone · 0.50
vendor_transferable0.06
Product, customer and vendor masters are governed centrally. Brightline's records are interleaved with Continental's in the same tables, and about 12% of customer records are shared accounts served by both. Deciding who owns a shared customer record is a co...
HRIS & payrollRun cost $2.2M human needed
day_one_critical0.96
tsa_exit_risk3.91 / 4 · 0.81
migration_complexity2.95 / 4 · 0.66
approachlift_and_shift · 0.72
data_gravity2.00 / 4 · 0.89
blocking_dependencyidentity · 0.50
vendor_transferable0.88
All 1,400 employees sit in Continental's Workday tenant with payroll run by their shared services centre across four countries. Payroll cannot miss a cycle; there is no tolerable failure mode. Workday quoted a standalone tenant at 20 weeks and confirmed his...
Plant time & attendanceRun cost $350k human needed
day_one_critical0.82
tsa_exit_risk2.82 / 4 · 0.42
migration_complexity2.18 / 4 · 0.48
approachrebuild · 0.70
data_gravity1.93 / 4 · 0.41
blocking_dependencyhris · 0.67
vendor_transferable0.14
Clock terminals at three plants feed Continental's timekeeping module, which feeds payroll. Two plants are unionised and the collective agreement specifies how time is recorded and disputed, so any change touches labour relations.
Data platform & warehouseRun cost $2.4M human needed
day_one_critical0.17
tsa_exit_risk3.70 / 4 · 0.62
migration_complexity3.81 / 4 · 0.71
approachrebuild · 0.58
data_gravity2.85 / 4 · 0.66
blocking_dependencyhris · 0.64
vendor_transferable0.18
Every reporting pipeline lands in Continental's lakehouse, with Brightline's tables co-mingled with other brands in shared schemas. Continental will grant a one-time bulk extract but no ongoing access after the TSA, and the extract has to be requested per s...
BI & management reportingRun cost $700k human needed
day_one_critical0.11
tsa_exit_risk1.98 / 4 · 0.59
migration_complexity2.02 / 4 · 0.53
approachrebuild · 0.72
data_gravity2.90 / 4 · 0.68
blocking_dependencydata_platform · 0.51
vendor_transferable0.77
About 900 dashboards, of which an internal audit found roughly 240 are opened more than once a month. Built on Continental's BI licence against their lakehouse.
Trade & customer CRMRun cost $1.3M human needed
day_one_critical0.89
tsa_exit_risk2.76 / 4 · 0.54
migration_complexity1.99 / 4 · 0.70
approachsaas_replace · 0.60
data_gravity2.05 / 4 · 0.73
blocking_dependencymdm · 0.51
vendor_transferable0.90
Salesforce org shared with three Continental brand lines. Brightline's accounts and trade-promotion records are separable by business unit, and Salesforce has a documented org-split service for exactly this scenario.
DTC storefrontRun cost $800k human needed
day_one_critical0.08
tsa_exit_risk0.23 / 4 · 0.65
migration_complexity1.00 / 4 · 0.77
approachlift_and_shift · 0.51
data_gravity0.98 / 4 · 0.67
blocking_dependencyerp · 0.82
vendor_transferable0.95
A Shopify Plus store already contracted directly in Brightline's own name after last year's DTC pilot. It reads inventory from the shared ERP but is otherwise self-contained.
Procurement & APRun cost $950k human needed
day_one_critical0.16
tsa_exit_risk2.98 / 4 · 0.56
migration_complexity2.88 / 4 · 0.45
approachrebuild · 0.52
data_gravity1.89 / 4 · 0.66
blocking_dependencyledger · 0.50
vendor_transferable0.79
Requisitions and invoice matching run in Continental's procurement suite against the shared ERP. Roughly 400 supplier contracts need renegotiating or novating regardless of the software question, and procurement has not started that work.
Quality management (FDA scope)Run cost $1.1M human needed
day_one_critical0.78
tsa_exit_risk3.55 / 4 · 0.50
migration_complexity3.73 / 4 · 0.63
approachrebuild · 0.32
data_gravity2.86 / 4 · 0.63
blocking_dependencyerp · 0.56
vendor_transferable0.20
Validated system holding batch records, deviations and CAPA history under FDA 21 CFR Part 11. Continental's validation package covers their instance, not a copy of it, so any new instance needs its own IQ/OQ/PQ. Regulatory counsel flagged that batch records...
Plant MES & historiansRun cost $1.8M human needed
day_one_critical0.88
tsa_exit_risk2.04 / 4 · 0.62
migration_complexity2.01 / 4 · 0.88
approachlift_and_shift · 0.57
data_gravity2.03 / 4 · 0.51
blocking_dependencynetwork · 0.69
vendor_transferable0.89
Line-level execution systems at three plants, largely self-contained on plant networks and already running on servers physically located at the plants that transfer with the facilities. Continental's involvement is limited to patching and remote support.
Lab information managementRun cost $600k human needed
day_one_critical0.15
tsa_exit_risk2.13 / 4 · 0.67
migration_complexity2.14 / 4 · 0.58
approachrebuild · 0.43
data_gravity2.08 / 4 · 0.63
blocking_dependencyqms · 0.68
vendor_transferable0.85
Two QC labs use Continental's LIMS. Validated, but far smaller in scope than the QMS. The vendor has an established migration path and Continental's lab team has done this once before for a divested plant.
Email & collaborationRun cost $1.2M human needed
day_one_critical0.98
tsa_exit_risk3.09 / 4 · 0.55
migration_complexity2.00 / 4 · 0.82
approachsaas_replace · 0.85
data_gravity2.92 / 4 · 0.57
blocking_dependencyidentity · 0.58
vendor_transferable0.06
Mailboxes, files and chat in Continental's tenant. Tenant-to-tenant migration is a well-trodden path with mature tooling, but every mailbox and the whole file estate has to move, and legal hold on about 40 mailboxes complicates the copy.
IT service deskRun cost $750k human needed
day_one_critical0.23
tsa_exit_risk3.84 / 4 · 0.73
migration_complexity1.90 / 4 · 0.62
approachsaas_replace · 0.52
data_gravity1.09 / 4 · 0.69
blocking_dependencyidentity · 0.61
vendor_transferable0.17
Continental's service desk handles Brightline tickets with 11 FTEs allocated. None of those people are transferring. Brightline needs both a tool and an org, and the org is the harder half.
Legacy MRP (Trenton plant)Run cost $240k human needed
day_one_critical0.14
tsa_exit_risk0.99 / 4 · 0.56
migration_complexity1.76 / 4 · 0.49
approachretire · 0.77
data_gravity1.19 / 4 · 0.65
blocking_dependencynone · 0.50
vendor_transferable0.04
An AS/400 application the Trenton plant kept after a 2009 acquisition. Continental has tried to decommission it twice. Its remaining function is a scheduling report that the plant manager says three people read, and the ERP has covered the same ground since...
Marketing automationRun cost $480k human needed
day_one_critical0.11
tsa_exit_risk2.32 / 4 · 0.40
migration_complexity2.72 / 4 · 0.40
approachsaas_replace · 0.63
data_gravity2.05 / 4 · 0.63
blocking_dependencycrm · 0.33
vendor_transferable0.88
Campaign tooling shared with Continental's marketing org. Brightline's consumer list is separable but consent records are held at the Continental entity level, and privacy counsel is still working through whether consent transfers with the brand in each jur...
Treasury & bankingRun cost $520k human needed
day_one_critical0.82
tsa_exit_risk3.81 / 4 · 0.69
migration_complexity3.80 / 4 · 0.69
approachsaas_replace · 0.69
data_gravity1.15 / 4 · 0.68
blocking_dependencyledger · 0.65
vendor_transferable0.10
Cash pooling, FX and bank connectivity all run through Continental's treasury function. Brightline needs its own bank relationships, entity-level accounts and payment rails from day one; you cannot pool cash with a company you no longer belong to.
Four prototypes of the TypeSafe AI / Jev System One contract applied to real decisions.
Source in businesses/jev-forecast/; ./jevctl.py dashboard --all
rebuilds every page. All companies, figures and hands are fictional.
Not affiliated with TypeSafe AI. Built September 2026.