Forecasts

Month-end spend projections and scenario modeling

Forecast horizon

Total Forecasted

$84,900

This Month projection

Total Budget

$67,500

This Month budget

Budget Variance

+25.8%

Projected overrun

Vendors at Risk

Forecasted over budget

Spend vs Budget vs Forecast by Vendor

Actual · Forecasted · Budget (dashed) — This Month

Vendor Forecast Detail

Sorted by risk level

Vendor	Category	MTD Spend	Forecasted	Budget	Variance	Confidence	Risk	Method
OpenAI	AI / LLM	$12,400	$19,800	$15,000	+32.0%	81%	critical	Weighted linear extrapolation (7-day trend)
Salesforce Agentforce	CRM / Agentic AI	$6,420	$13,100	$8,000	+63.8%	68%	critical	Flex Credit consumption rate (post-spike trend)
Anthropic Claude	AI / LLM	$8,900	$14,600	$12,000	+21.7%	77%	high	Weighted linear extrapolation (7-day trend)
AWS	Cloud Infrastructure	$10,700	$17,300	$15,000	+15.3%	74%	high	Service-level extrapolation + anomaly adjustment
Snowflake	Data Warehouse	$5,600	$7,800	$6,000	+30.0%	79%	high	Credit consumption rate (warehouse-level)
Datadog	Observability	$4,200	$6,000	$5,000	+20.0%	85%	medium	Linear extrapolation (30-day baseline)
Google Vertex AI	AI / LLM	$2,900	$4,300	$4,000	+7.5%	75%	medium	Linear extrapolation (7-day trend)
Twilio	Messaging	$1,300	$2,000	$2,500	-20.0%	88%	low	Linear extrapolation (30-day baseline)

What-If Scenarios

OpenAI usage continues at current rate

critical

If GPT-4.1 token consumption stays at the current 7-day trajectory through month end

Projected month-end

$19,800

Budget impact

+$4,800

Assumption: Current burn rate of $748/day maintained for remaining 3 days

Suggested action

Enable prompt caching and review endpoint usage with Product Engineering today

Claude output token usage continues increasing

high

If Anthropic Claude output token growth trend continues through month end

Projected month-end

$14,600

Budget impact

+$2,600

Assumption: 5% week-over-week output token growth rate maintained

Suggested action

Enable prompt caching on /api/contracts/analyze and set max_tokens limits

AWS workload returns to baseline

low

If the Lambda PDF processor batch schedule is corrected and invocations normalize

Projected month-end

$13,200

Budget impact

-$1,800

Assumption: Lambda invocations return to 45,000/day by Apr 30

Suggested action

Confirm batch schedule fix with Platform Engineering and monitor for 24h

Agentforce actions increase another 20%

critical

If support automation continues expanding beyond the initial rule change

Projected month-end

$15,720

Budget impact

+$7,720

Assumption: Flex Credit consumption grows an additional 20% from current elevated level

Suggested action

Set a Flex Credit daily cap in Salesforce Digital Wallet and audit automation rules

AI platform usage grows 15% week over week

high

Compound growth scenario if OpenAI, Claude, and Vertex AI all sustain 15% WoW growth

Projected month-end

$52,000

Budget impact

-$3,000

Assumption: 15% weekly compound growth across all AI/LLM vendors

Suggested action

Establish per-feature token budgets and add caching across all AI endpoints