An advisory firm makes recommendations. Dimension designs, builds, and iterates the systems that execute on them. What follows: featured engagements, the operational categories the firm architects, and a statement on the velocity at which the firm moves.
A multinational conglomerate operates four distinct subsidiary businesses across four continents — US retail, LatAm agribusiness, EU financial services, Japanese manufacturing. Each runs on its own legacy stack, shaped by its own industry and its own regulatory environment. The systems work individually. The problem is that none of them communicate a synthesized read of operations to the parent business as a whole. Analysts re-key numbers across reports; finance teams reconcile definitions across regions; executive assistants stitch slide decks together the night before a board meeting. By the time data reaches the boardroom it is already stale, and the synthesis is editorial. Bad decisions are made in any business operating from data that does not represent the entire picture.
The firm's charge was to replace the night-before-the-board ritual with a system that produces a defensible, current operating view on demand. The work began with the executive teams across each subsidiary — not with a technology selection. The firm sat with each business unit to identify which data, reports, and metrics must roll up to the parent's global view, and which stay local. That negotiation, before any code was written, defined the entire downstream architecture. The unlock was Salesforce — not as a CRM, but as a composable platform flexible enough to absorb the differences between subsidiaries while delivering harmonization, visualization, distribution, and intelligence as a coherent native suite. Each subsidiary runs its own Salesforce organization shaped to its industry, formed from the legacy systems, which in some cases are still operating but now feeding up. Regional user interfaces moved headless — bespoke front-ends matching local workflow conventions, local language, and local regulatory display requirements, calling into the platform underneath.
Subsidiary operational data flows into an Amazon-hosted data lake with region-prefixed storage to preserve data sovereignty obligations under GDPR, APPI, LGPD, and US sectoral rules. Salesforce Data Cloud federates from the lake using zero-copy — the executive layer queries the data where it lives rather than duplicating it across systems. That is the architectural lever that makes the platform viable at scale: petabytes of subsidiary operational data without petabytes of duplication, with governance enforced at the storage layer where it belongs. The harder problem is semantic. The retail subsidiary's "customer" is a consumer with loyalty data; the agribusiness's is an off-taker on a multi-year contract; the financial services subsidiary's is a regulated counterparty; the manufacturer's is a B2B industrial purchaser. Rolling those into a single global customer count requires explicit semantic work — Data Lake Objects per subsidiary mapped into Data Model Objects that serve as the common executive vocabulary. A central data council, drawn from each subsidiary plus the parent's office of the CEO, owns those metric definitions and resolves subsidiary objections at the governance layer rather than the engineering one. The harmonization layer encodes negotiated truth, not imposed truth.
The platform launched approximately eighteen months ago with the core architecture: subsidiaries feeding the lake, Data Cloud harmonization, Tableau as the executive visualization layer. Within the first year, consumption shifted — executives stopped opening dashboards and started consuming insights inside Slack, where their working day already lived. Tableau Pulse delivers targeted proactive insights to channel; subscriptions deliver scheduled briefings. Slack messages that deliver real metrics are welcomed; messages that demand attention without earning it are unsubscribed. The platform is now in its third phase: Agentforce, the agentic layer grounded in the harmonized Data Cloud model. Agents reason across the four subsidiary domains and surface unified insights — anomalies, trends, exceptions — back into the same Slack channels and the regional headless interfaces operators already use. Where governance supports it, agents take action back into the source Salesforce organizations. The parent executive team now operates from a single, defensible picture of the conglomerate — refreshed continuously, harmonized across four distinct businesses, accessible through the surfaces executives already use.
A national operating network grew from zero to seventy-five thousand members across roughly forty cities in a handful of years. That growth pattern — fast acquisition, geographic distribution, federated local operations — is one of the hardest configurations to architect well. Each city operates with its own programming, its own local leadership, and its own community. The challenge is to give those local operations real autonomy while running everything on a single data spine that headquarters can actually see, govern, and report on.
The firm built the complete operating platform that runs the network end-to-end. New member signups flow from the public website into Salesforce, where a record-of-truth is established and reconciled. Iterable orchestrates the network's full marketing automation — onboarding journeys, lifecycle communications, behavioral triggers — all wired to member activity in Salesforce. Each new member is automatically associated with their nearest active city — geofenced within a hundred to two hundred miles — and provisioned into that city's local operational stack. Each city operates from its own Slack workspace, each tied to its own Salesforce Account and Parent Campaign so that all local activity rolls up cleanly into the firm-wide data model. Tableau handles visualization at the leadership layer; Ramp handles event-level expense management. Every platform reads from the unified Salesforce truth source, so reporting and reconciliation are internally consistent.
The hardest operational problem in any federated organization is events. Local events generate personal information, venue coordination, public-facing registration pages, attendee tracking, and post-event documentation — and every commercial event platform on the market solves one piece poorly and the others worse. So the firm built the bridge. Local leaders create events through a structured Slack form: ten to twelve questions tailored to the event type. On submission, the approval team is notified in Slack and the event record lands in Salesforce on a custom Event Approval object. Once approved, a single button creates a Salesforce Campaign that syncs to the corporate WordPress site, which generates the public event page automatically. From there, attendee signup flows back into Salesforce, confirmations are triggered, reminders are scheduled. After the event, the leader files a structured After Event Review that closes the loop. The local leader runs the event; the platform runs the operations.
The same Slack-form pattern extends to member communications. A leader composes a message in a structured Slack form — subject, body, send timing. On submission, the content is templated to brand standards inside Iterable and dispatched to the leader's local membership, segmented from the unified Salesforce data. The leader never leaves Slack; the marketing automation layer handles formatting, segmentation, deliverability, and tracking. That is the architectural principle running through both workflows: Slack is the composition surface for local leaders; the enterprise platforms beneath it execute the operation.
A presidential primary is, in operational terms, a Fortune-500-scale enterprise that has to be built, run, and dismantled inside a fifteen-month window. There is no Q2 recovery, no second launch, no soft release. The operation either works on day one and every day after, or it doesn't.
For the 2016 Presidential Primary, the firm — then operating as CFB Strategies — architected and built the campaign's complete fundraising and bundler engagement platform. The system reconciled more than twenty-four distinct funding sources daily, allocated incoming dollars in real time so leadership could see, on any given morning, exactly what the day before had produced and where it came from, and routed contributions through compliant payment infrastructure under federal election law.
The platform also handled the campaign's grassroots data infrastructure. Volunteer signups flowed in from the public website across every state and territory, captured in Salesforce with the geographic, demographic, and skills data needed to route each volunteer into the right local operation. By the end of the cycle, the system carried more than three hundred thousand active grassroots volunteers — every state covered, every region mapped, every contact addressable at the data layer.
The platform's distinguishing feature was the bundler experience. Each high-value bundler was issued a private page within the system — a personal dashboard tracking their book of donations, their progression through the program's tiers, and the recognition milestones they had earned. The experience was gamified. Awards, tier advancement, and visible standings turned a process that is traditionally opaque into a competitive one.
A working system is rarely one platform. The client arrives with a destination — a launch, a regulatory deadline, a transformation, a program that has to be live inside a defined window — and no architecture between where they are and where they need to be.
The first work is the orchestration of the stack itself. The firm operates across the major enterprise architectures — Microsoft, AWS-native, Snowflake, Databricks, SAS, and any data-lake configuration. A frequent Dimension default is AWS or Data 360 as the cloud foundation, Salesforce as the system of record, Slack as the operational nervous system, Tableau as the analytic lens, and a marketing automation platform at the perimeter, each chosen per engagement. The architectural pattern is independent of which vendor holds which role. The integrations between platforms — the data flows, the auth boundaries, the workflow handoffs, the reporting contracts — are the actual product. The platforms are the materials; the orchestration is the building.
A Dimension specialty worth surfacing explicitly: the firm builds custom user-facing interfaces that read against Salesforce as the underlying data spine but present to the end user as a bespoke product — not as a CRM that has been skinned. Most Salesforce deployments look like Salesforce; the UI fingerprint is unmistakable to anyone who has seen one. Salesforce recently featured this approach publicly as headless; the firm has been building this way for years. With the advancements of AI and LLMs, headless implementations have become a common pattern. The end user never recognizes the implementation as a Salesforce build — that is the point.
By the time a Dimension engagement reaches production, those platforms read as one observable, owned, modifiable surface — not five vendors talking past each other, not a stack of dashboards no one can correlate, not a workflow held together by a junior employee's spreadsheet. The categories that follow describe what the orchestration is in service of.
Although each build is different, the work often starts out within one of these operational categories. Each is built end-to-end, owned by the firm, and designed as an integrated system rather than a collection of bolted-together tools.
Every category above now has a model layer. Marketing operations runs on classification and personalized content generation. Fundraising layers in predictive scoring and donor intelligence. Regulatory work uses document intelligence and audit-grade retrieval. Government affairs deploys constituent-message triage and legislative monitoring at scale. The firm builds those capabilities directly into the platform the client already operates — not as a parallel AI tool the staff has to learn separately, not as a chatbox bolted onto the side of an existing workflow.
The model itself is chosen per engagement — Claude, GPT, Gemini, or open-weight — selected on the task, the data-residency constraints, the latency envelope, and the cost per inference. The integration target is the system of record: Salesforce, Slack, the data warehouse, the marketing layer at the perimeter. The work spans classification, document intelligence, structured output generation, agentic workflows, retrieval against proprietary corpora, and automated triage at volume. The model is replaceable; the integration is the product.
Above the model layer sits the agent layer. The firm builds autonomous agents that operate across the integrated stack. For Salesforce-ecosystem engagements, Agentforce covers the surface natively — Sales Cloud, Service Cloud, Data 360, Tableau, MuleSoft, and Slack (which Salesforce owns, so the agent reach into team conversation runs deep). Atlas, the reasoning engine inside Agentforce, is model-agnostic — it works with Einstein, Anthropic, OpenAI, or any other LLM the engagement calls for. For engagements that extend beyond the Salesforce ecosystem — into Iterable, Ramp, custom applications, third-party data sources — the firm builds custom agent frameworks that orchestrate across the broader stack, increasingly through MCP-based patterns. The firm builds for both.
The firm operates with these tools in daily practice. That practice produces operator-level fluency with the models — what they can actually do, what they cost in tokens and latency, where they fail in production. AI deployed as working infrastructure, judged on whether it ships and survives operations, not on whether it demos well.