Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
AI's Valley of Despair: Why Process Redesign Beats Faster Models - OpenSmartRoute
AI's Valley of Despair: Why Process Redesign Beats Faster Models
Companies face AI disappointment because speed gains are blocked by slow manual steps. Executives see low earnings growth while costs rise due to unoptimized workflows.
Key points
Only 15% of AI decision-makers saw earnings increase last year.
Firm business process redesign takes 90 to 120 days to implement.
Organizations should expect firm ROI data within about six months.
Companies must track cost per outcome instead of average costs.
Why it matters: Teams must align AI spend with measurable business outcomes like revenue or risk reduction.
By OpenSmartRoute editorial · written through the router by writer-small
From SiliconANGLE AI - “How to avoid AI’s ‘valley of despair’”
The Valley of Despair Explained - What Amdahl's Law means for AI adoption
Amdahl's Law states that improving a single part of a system has a limit on how much it speeds up the whole. Even if one step becomes exponentially faster, other steps impose a ceiling unless they also accelerate. This concept explains why companies feel disappointed with artificial intelligence adoption. An AI assistant can draft a proposal in seconds, but it waits for days for reviews.
Similarly, a coding tool produces software in minutes, yet testing takes weeks. Companies accelerated specific tasks without reengineering the surrounding systems. This gap between inflated expectations and modest returns is called AI's "valley of despair." Executives see low earnings growth while costs rise due to unoptimized workflows.
The Reality of AI Returns - Low earnings growth and limited value mapping
Analyst firm Forrester Research Inc. found that only 15% of AI decision-makers reported an increase in earnings over the past 12 months. Fewer than one-third could clearly map the value of AI back to bottom-line changes. This skepticism is healthy for AI investments. AI investments should face the same scrutiny as any other capital expense.
Leaders need to be clear on expected returns and timelines. They must show whether their investment delivers measurable business results. Small, localized productivity improvements are not game changers. Giving employees access to general-purpose AI tools improves discrete tasks.
However, this value is diffuse and disconnected from business performance metrics. This is Amdahl's Law playing out at scale. AI does have business-transforming potential, but isolated gains do not transform the business.
Why Automation Fails - The gap between speed and business impact
Organizations often seek to automate what they already do rather than reinvent it. A more mature approach requires reassessing and reengineering how the business operates. Applying AI creates fundamentally new operating processes.
Whereas automation asks how existing tasks can be performed faster, reengineering asks whether a task should exist. Time saved does not automatically become revenue or margin. What does AI ROI actually look like for an organization? Generally speaking, organizations target positive, measurable change in revenue, efficiency, or risk.
SandboxAQ launched Switch Trust and Switch ROI to help companies manage risk and measure value for their AI agents. These tools extend the open-source Switch Rooms platform to provide security checks and cost tracking.
The Agentic AI Foundation now maintains ten open-source projects. It includes standardized APIs, skill formats, and sandbox environments donated by major companies.
A quantifiable business goal might be to reduce customer churn. It might shorten the lead-to-proposal cycle or improve forecast accuracy. It might detect and manage fraud faster. These are the areas where true value appears.
The Path to Real Value - Reengineering processes instead of just automating
Enterprises should reasonably expect firm ROI data within about six months. Leading indicators like scenario testing can be delivered quickly. Large-scale, immediate returns are unrealistic for most companies. A thoughtful business process redesign takes 90 to 120 days to design and implement.
This period includes acceptance testing and focused change management. These initiatives succeed when people accountable for results deliver a specific outcome. They do not simply achieve a technology go-live. Costs deserve more sophisticated treatment because expenses vary drastically.
The expense of agentic processes depends on use case complexity and model calls. Companies should measure the distribution of cost per outcome. They should not rely on a pre-deployment average. Early adopters find that monitoring cost per case is crucial.
They must refine their strategy to keep numbers within an acceptable range. This helps successfully redesign business processes with AI. Treat AI as a series of targeted business transformations. Pair clear outcomes with disciplined execution.
Measuring True ROI - Tracking revenue, efficiency, and risk metrics
Take a closer look at business processes within the organization. Examine how work actually happens, not just how a procedure "should" happen. Identify bottlenecks, handoffs, approvals, and exceptions before applying AI. Expect to find substantial differences between documented processes and reality.
Tie every initiative to a clear outcome. Name the revenue, efficiency, or risk metric the AI initiative impacts. Establish a baseline and give a business leader ownership of the result. Set a hard timeline for impact data.
Design pilots for production. AI pilots should test the path to sustained use. They should not merely demonstrate impressive capabilities under controlled conditions. Include operational data, integration, governance, security, and change-management requirements from the beginning.
Measure economics, not activity. Track metrics such as cycle time, quality, adoption, and cost per outcome. Do not track seats, prompts, or demos. For agentic systems, monitor both the range and the average of operating costs. Complex cases can consume far more resources than simpler ones.
Review and adjust regularly. Decide whether to move forward, narrow the scope, redesign the workflow, or stop. Exploratory work won't always succeed, but it should produce evidence. This evidence improves the next capital allocation.
Why it Matters
Companies that understand the value of investing time and capital to rethink processes can turn AI spend into tangible outcomes. AI's "valley of despair" is not inevitable. Executives can avoid disappointment by focusing on process redesign. They must focus on reengineering business processes instead of just buying faster models.
This approach ensures that AI investments deliver measurable business results. It aligns technical capabilities with actual revenue and efficiency gains. It prevents the waste of capital on isolated speed improvements.
How OpenSmartRoute Helps
Teams that route requests through OpenSmartRoute gain precise control over which model handles each task. The router sends each request to the best-fit model, agent, tool, or skill from a defined catalogue. It scores every candidate on quality, cost, speed, and safety. The team sets the weights per request to prioritize these factors.
Hard rules pin specific requests to specific models. For example, text with personal data stays on an on-premises model. A region or a cost cap is never crossed. The system learns from outcomes, so a model that answers well gets more traffic. A model that fails gets less traffic.
A new model is just one catalogue entry and competes on the next request. Nothing else changes in the app. The hosted platform keeps a models catalogue with prices and public rankings built from real traffic. A savings ledger shows what each routed request cost next to what the most expensive model would have cost.
An input guard spots prompt injection and personal data before a request leaves. 'osr eval' measures routing accuracy on the team's own prompts and can fail a build when it drops. This setup directly addresses the cost and quality issues described in the valley of despair.
How it compares
Amdahl's Law explains why speeding up one part of a system does not speed up the whole system. An AI tool might write code in minutes, but human review still takes days. This delay creates a bottleneck that limits total speed. Companies often buy tools to automate tasks without changing the surrounding workflow. They expect the tool to fix the entire process on its own. This approach leads to frustration when results are slower than expected. The "valley of despair" comes from this mismatch between hype and reality.
Old methods focused on automating existing steps to save time. New strategies focus on reengineering the entire business process. Reengineering asks if a task should exist at all before trying to automate it. This shift moves the goal from faster execution to creating new value. The technology remains the same, but the business logic changes completely. Teams must redesign how work flows before adding AI tools. This ensures that every step contributes to the final outcome.
The previous approach treated AI as a magic button for immediate results. The new approach treats AI as a series of targeted business transformations. Outcomes matter more than the technology itself. Leaders must define specific metrics like revenue or risk reduction. They need to track these metrics over six months to see real ROI. Early promises of quick returns are often unrealistic for large-scale changes. Patience and disciplined execution are required for success.
Some teams still try to automate everything without rethinking the process. They fail to measure cost per outcome or track financial impact. They rely on average costs instead of monitoring distributions. This ignores how complex cases consume more resources than simple ones. The result is a budget that grows while value stays flat. Teams need to stop measuring activity and start measuring economics. They must track cycle time and quality alongside financial data.
The core difference lies in who owns the result. Old models gave technology teams ownership of the deployment. New models give business leaders ownership of the outcome. This ensures accountability for the final business result. Exploratory work must produce evidence for the next capital allocation. Teams cannot simply go live and hope for the best. They must continuously review and adjust their strategies based on data.
Questions this leaves open
The source text does not specify which AI models are best for reengineering processes. It mentions monitoring costs and quality but does not name specific vendors. Readers cannot know if a particular model will handle complex cases better. They need to check benchmark results and real-world evaluations before buying. Open weights might offer more transparency than closed models for this purpose.
The article assumes teams have the budget to redesign workflows. It does not explain how to fund these expensive reengineering projects. Smaller companies might struggle to afford the time and capital required. They need to know if there are cheaper ways to test these ideas. Pilot programs can be delivered quickly, but scaling them requires significant investment.
There is no clear timeline for when ROI becomes visible for every company. The text says six months is reasonable, but this varies by industry. Some sectors might need longer to see measurable changes in revenue. Readers should check their specific industry standards and regulatory requirements. Safety and compliance might extend the time needed to deploy new processes.
The source does not detail how to handle organizational resistance to change. Change management is listed as a requirement, but no specific tactics are given. Leaders might fear losing control or facing new job roles. Teams need strategies to communicate the value of reengineering to all staff. Training programs could help employees adapt to new workflows and tools.
It remains unclear how to balance speed and safety during the redesign phase. Agentic systems can consume far more resources on complex cases. Teams must decide how much risk they are willing to take on early pilots. They need to set strict guardrails before allowing unmonitored testing. Security and governance requirements must be met before any data leaves the system.
Readers might wonder how to validate the "cost per outcome" metric. They need to define what constitutes a single outcome in their business. Is it one customer interaction or one product sale? The definition changes the cost calculation significantly. Teams must agree on a standard unit of measure before tracking costs.
The text implies that reengineering is always necessary for success. It does not explain when incremental automation is still the right choice. Some tasks might be simple enough to automate without full redesign. Teams need to know when to stop reengineering and start automating. A clear boundary between the two approaches helps allocate resources efficiently.
Finally, the source does not address how to maintain momentum after the initial success. What happens when the first pilot shows positive results? Teams might get complacent and stop measuring outcomes. Continuous improvement requires a culture that values data over ego. Leaders must keep asking hard questions about every new initiative.
What to Do
Start by taking a closer look at business processes. Examine how work actually happens within the organization. Identify bottlenecks, handoffs, approvals, and exceptions before deciding where to apply AI. Expect to find differences between documented processes and how work gets done.
Tie every initiative to a clear outcome. Name the revenue, efficiency, or risk metric the AI initiative is intended to impact. Establish a baseline and give a business leader ownership of the result. Set a hard timeline for impact data.
Design pilots for production immediately. AI pilots should test the path to sustained use, not just demonstrate capabilities. Include operational data, integration, governance, security, and change-management requirements from the beginning.
Measure economics, not activity. Track metrics such as cycle time, quality, adoption, cost per outcome, and financial impact. For agentic systems, monitor both the range and the average of operating costs.
Review and adjust regularly. Decide whether to move forward, narrow the scope, redesign the workflow, or stop. Exploratory work should continually produce evidence that improves the next capital allocation.
Inception AI launched Mercury 3, which is up to eight times faster than GPT-6 Luna and Claude Haiku 5.5. The model uses a new diffusion approach to generate text at 1,068 tokens per second.
