ChatGPT Without ROI: What to Deploy Next — Explained by Softprom, OpenAI and Superhuman Partner
News | 28.08.2026
“We handed out the licenses six months ago. I still haven’t seen a single ‘we got faster’ report.” If this sounds like your company, you are not an exception — you are the statistic: according to MIT research (Project NANDA, 2025), 95% of corporate GenAI pilots deliver no measurable P&L impact¹. Softprom — a value-added IT distributor since 1999, an OpenAI partner (OpenAI SMB Channel Partner) and supplier of Superhuman solutions across Central and Eastern Europe, the Caucasus, and Central Asia — explains why access to ChatGPT alone does not raise productivity, and what to deploy as the second step.
The Short Answer
Productivity did not grow — not because the model is bad, but because AI remained a separate tab instead of becoming part of the work processes. MIT researchers name the “learning gap” as the main cause of failures: the tools are not embedded in real workflows and do not learn from them¹. The second step after “hand out ChatGPT” is to close the gap between AI and the place of work: connect corporate context (ChatGPT Business / Enterprise connectors from OpenAI) and add a built-in layer for every employee’s daily tasks (the Superhuman Suite).
Why “Everyone Has ChatGPT” ≠ “Everyone Works Faster”
AI demands a habit change: go to a tab, write a prompt. The fix — tools that work in the input field without changing habits: Grammarly and Superhuman Go from Superhuman are already where people write.
A public chat without corporate context cannot know them. The fix — ChatGPT Business / Enterprise with connectors to the company’s repositories and knowledge base.
Fix 2–3 metrics before rollout (customer response time, time per typical document, share of emails sent without a manager’s edits) and measure monthly — otherwise ROI stays invisible even when it exists.
Shadow AI disappears when the corporate tool is more convenient than the personal one: single sign-on, administration, data policies in the business editions of both vendors.
The Second Step: A 90-Day Plan
-
Days 1–30. Diagnostics and metrics
Measure where the team actually loses time (email, documents, customer replies), record the baseline values. Pick 2–3 pilot teams with writing-heavy work — sales, support, marketing.
-
Days 31–60. The built-in layer
Roll out Grammarly Business / Superhuman Go to the pilot teams: the tool works in 100% of texts automatically, so the effect is visible in weeks. In parallel — connect ChatGPT Business connectors to corporate sources for analytical tasks.
-
Days 61–90. Measure and scale
Compare the metrics against the baseline, collect team cases, roll out to the rest of the company. The before/after comparison is the ROI report that was missing.
Companies in 33 countries build this stack through Softprom: Superhuman licenses directly, OpenAI solutions via the OpenAI SMB Channel Partner referral program; with support for the pilot and rollout metrics.
FAQ
According to MIT (NANDA, 2025), 95% of corporate GenAI pilots deliver no measurable ROI — mostly because AI is not embedded in work processes and has no corporate context, not because of model quality.
Two additions: ChatGPT Business / Enterprise connectors (OpenAI) to corporate data — and a built-in layer for every employee’s daily writing and actions (Superhuman: Grammarly, Superhuman Mail, Superhuman Go).
Record baseline metrics before rollout (customer response time, time per document, share of texts without edits) and compare monthly. Without baseline values, ROI is invisible even where it exists.
A realistic cycle is 90 days: diagnostics and metrics (30), a pilot on writing-heavy teams (30), measurement and scaling (30).
Softprom — a value-added IT distributor since 1999: supply of Superhuman solutions and connection to OpenAI solutions under the OpenAI SMB Channel Partner program in 33 countries, with pilot and rollout support.
¹ MIT Project NANDA, “The GenAI Divide: State of AI in Business 2025”: 95% of corporate GenAI pilots show no measurable P&L impact; the key cause is the “learning gap” — tools not embedded in workflows.
Share:
