3 of 15

Scope of Services

Soft methods & problem structuring
Historical trends & data analysis
Modelling, simulation & wargaming
R&D, concepting, experimentation
Capability planning
Acquisition

P3M Design, implementation & Strategic Advice
P3M Maturity Assessments
Change Management and transition support
PMO Improvement initiatives
End to end project management & project leadership
Bid Management
Collaboration leadership

Product, service & enterprise level
Tailorable & scalable lifecycle solutions
Requirements, Architecting
Systems Analysis
Integration Verification Validation & Testing
Sustainment
Disposal & Replacement

Business Case support
5 Case Model
HMT Approvals
Cost-benefit, cost-effectiveness and Option selection
Value for Money cases
Financial estimation
Risk management
Socio-economic benefits

Whole life cost models

Should-/Could-cost
Cost-Benefit
Value-For-Money

Cost estimation

Acquisition
Through-life
All domains, platforms & DLODS

Investment Appraisals
Risk Analysis
Programme Mgt
Business Cases

Cost

Analysis

Approvals

Systems

Engineering

OA / OR

P3M

Delivering clarity from complexity

4 of 15

Our Clients & Partners – trade references on request

www.sirius-analysis.co.uk

Delivering clarity from complexity

5 of 15

What Does the Future Hold for

Cost Engineering?

A strategic outlook for practitioners, organisations,

and the profession at large

Software Cost Estimation & Beyond | Sirius Analysis Ltd

70%

Projects exceed

cost estimates

$2T+

Wasted annually on

failed IT projects

<40%

Estimates within 10%

of actual cost

6 of 15

AGENDA

State of Cost Engineering Today

Maturity, persistent overruns, changing environment

Disruption Forces Reshaping the Discipline

AI, Cloud, Agile, ESG, Distributed Teams

AI in Cost Estimation

ML methods, LLMs, limits and governance

Reinvention of Classical Methods

Parametric models, analogy-based, reference class forecasting

Agile, Uncertainty & Probabilistic Framing

Rolling wave, Monte Carlo, Agile EVM

Digital Twins & Real-Time Forecasting

Continuous programme cost control

Sustainability & the Green Cost Engineer

Carbon as a cost dimension, regulatory risk

Future Skills, Roles & Bold Predictions

2030 outlook and call to action

7 of 15

01 | STATE OF COST ENGINEERING TODAY

Established Methods

Parametric Modelling — COCOMO III, SEER-SEM, SLIM
Analogy-Based Estimation using project repositories
Expert Judgement — Planning Poker, Delphi, 3-Point
Function Point Analysis (Albrecht) & COSMIC
Bottom-Up Effort Estimation

Why overruns persist

Scope instability & optimism bias
Inadequate historical project data
Poor requirements quality
Underestimation of complexity

"The challenge is no longer whether we have enough estimation methods. It is whether we have the organisational will, data discipline, and adaptive capability to apply them well."

SPEAKER NOTES — Slide 3: State of the Discipline��The profession has four decades of accumulated method: COCOMO, Function Points, SEER-SEM, expert judgement. These are not obsolete — but they were built for a different era.��The bar chart illustrates the persistent problem: fewer than 40% of estimates land within 10% of actual cost. Despite four decades of methodological development, overruns remain endemic.��Five root causes recur across programme post-mortems: scope instability, optimism bias, data gaps, poor requirements, and a chronic tendency to underestimate complexity.��The environment is now changing faster than most practitioners' mental models of it. The question is no longer whether better methods exist — it is whether organisations have the will and the data infrastructure to use them.

8 of 15

02 | FIVE FORCES RESHAPING THE DISCIPLINE

AI & Automation

Generative AI rewrites, tests and documents code. Existing effort models need urgent recalibration.

Cloud-Native & SaaS

CAPEX → OPEX shift. Variable cost curves; FinOps collaboration essential for TCO modelling.

Agile & Hybrid Delivery

Fixed-price estimates clash with incremental discovery. Probabilistic forecasting is the answer.

ESG & Regulatory Pressure

Carbon costs, energy consumption and compliance costs are now material financial risks.

Distributed Teams

Offshore cost arbitrage eroded by communication overhead. Location-adjusted productivity models needed.

SPEAKER NOTES — Slide 4: Five Disruption Forces��Each of these forces is individually significant. Together, they constitute a fundamental reshaping of the practice environment.��1. AI & Automation: Generative AI tools are already writing, testing and documenting production code. Every parametric model calibrated on human-written code now has a recalibration problem. The productivity multiplier assumptions built into COCOMO and SEER-SEM need urgent revision.��2. Cloud-Native & SaaS: The shift from capital to operational expenditure transforms TCO modelling. Infrastructure costs are now variable, non-linear and consumption-based — not fixed assets you can schedule depreciation against.��3. Agile & Hybrid Delivery: Finance functions still demand upfront commitments. Delivery teams work in two-week sprints with deliberate uncertainty. The cost engineer must bridge these worlds with probabilistic forecasts, not pseudo-precise point estimates.��4. ESG & Regulatory Pressure: Carbon pricing, the EU Taxonomy, SECR, and ISSB standards are creating new compliance cost dimensions. These will become standard line items in business cases within this decade.��5. Distributed Teams: Simple offshore cost arbitrage is more complex than it looks once communication overhead, rework and coordination costs are properly modelled.

9 of 15

03 | ARTIFICIAL INTELLIGENCE IN COST ESTIMATION

Opportunities

Ensemble ML Models

Random Forest, Gradient Boosting & XGBoost outperform regression models on curated data

NLP Scope Extraction

LLMs extract functional entities from SOW/requirements text to automate Function Point preparation

Analogy Vector Search

Embedding models enable semantically rich similarity matching across historical project repositories

Speed & Consistency

AI first-cut estimates generated in seconds — human cost engineers provide calibration and governance

Risks & Governance

Training Data Quality

Models trained on inconsistent data will systematically mislead — GIGO applies absolutely

Explainability Imperative

In defence, healthcare & public sector: black-box estimates are not acceptable for programme approval

Model Drift

As technology stacks evolve, statistical relationships become stale — continuous recalibration is mandatory

Algorithmic Bias

Historical data encodes past inequities — regular bias audits must be built into governance processes

"AI-powered estimation models are not a replacement for the cost engineer's expertise. They are a powerful tool in the expert's hands — and a dangerous one without it."

SPEAKER NOTES — Slide 5: AI in Cost Estimation��This is the most significant methodological development since parametric modelling arrived in the 1980s. But the enthusiasm must be matched with clear-eyed realism.��Opportunities:�• Ensemble methods (Random Forest, XGBoost) consistently outperform parametric regression on well-curated datasets�• LLMs can extract scope from requirements documents and assist with Function Point preparation — reducing manual effort and improving consistency�• Vector similarity search transforms analogy-based estimation: semantic matching rather than surface attribute comparison�• AI produces first estimates at machine speed — freeing cost engineers to focus on calibration, risk and governance��Risks:�• Garbage in, garbage out: training data quality is the single biggest constraint�• In regulated environments (MoD, NHS, nuclear), explainability is a professional and legal obligation — XAI methods like SHAP and LIME are tools, not optional extras�• Model drift is an underappreciated risk: yesterday's model calibrated on yesterday's technology stack�• Bias audits must be scheduled and documented��The key message: the cost engineer who understands AI's capabilities AND limitations is invaluable. The one who trusts it uncritically is a liability.

10 of 15

04 | REINVENTION OF CLASSICAL METHODS

Parametric Models

COCOMO III · SEER-SEM · SLIM

Powerful theoretical foundations but calibrated in the waterfall era. The future lies in Bayesian calibration — updating parameters using organisational historical data as a prior. Integration with ALM tools (Jira, Azure DevOps) automates size and complexity metric extraction.

Analogy-Based Estimation

Vector Similarity & Ensemble Analogy

Intuitively appealing but historically weakened by imprecise similarity definition. LLM embedding models encode project descriptions as high-dimensional vectors, enabling semantically rich matching. Ensemble analogy — blending top-N similar projects — yields probabilistic estimates with natural confidence intervals.

Reference Class Forecasting

Kahneman & Tversky → HM Treasury Green Book

RCF begins with the outside view: use the outcome distribution of comparable historical projects as the forecast baseline before adjusting for project-specific factors. Required for major UK public sector projects under HM Treasury Green Book guidance. Powerful antidote to optimism bias.

SPEAKER NOTES — Slide 6: Reinventing Classical Methods��Reports of the death of classical methods are premature. What IS true is that they need significant reinvention.��Parametric Models: COCOMO III, SEER-SEM and SLIM are not obsolete. Their weakness is that default parameters reflect a waterfall, monolithic, co-located project world. The innovation needed is Bayesian calibration — feeding organisational project history as a prior to update model parameters. Pair this with ALM tool integration to automate metric extraction.��Analogy-Based Estimation: The conceptual appeal is strong; the historical weakness was defining 'similarity' rigorously. LLM embeddings solve this: project descriptions encoded as vectors, semantically matched against historical repositories. Ensemble analogy blends multiple top matches to give probabilistic outputs with confidence intervals.��Reference Class Forecasting: Rooted in Nobel Prize-winning behavioural economics research. The Green Book now explicitly recommends it for major UK public sector programmes. It forces estimators out of the 'inside view' (optimism about this specific project) into the 'outside view' (what actually happens to projects like this one). Particularly valuable for challenging optimism bias in early-stage estimates.

11 of 15

05 | AGILE, UNCERTAINTY & PROBABILISTIC FRAMING

From Point Estimates to Probability Distributions

The traditional model — a single-point estimate produced at initiation and defended to completion — is fundamentally incompatible with agile delivery.��Probabilistic Cost Framing: Output a distribution — P50, P80, P90 — not a point. Monte Carlo simulation using historical data provides rigorous, auditable distributions.��Rolling Wave Estimation: Near-term work estimated in fine-grain detail using team velocity. Horizon work estimated at coarser fidelity commensurate with available knowledge.��Agile EVM: EVM metrics calculated from sprint velocity data. Financial governance visibility without imposing waterfall constraints.

"False precision is the enemy of good decisions. A probabilistic estimate that is honest about uncertainty is worth more than a point estimate that creates an illusion of certainty."

SPEAKER NOTES — Slide 7: Agile, Uncertainty & Probabilistic Framing��The tension between agile and cost engineering is not inherent — it is a matter of implementation.��The root problem: finance functions demand firm upfront commitments. Agile methods deliberately embrace uncertainty. The cost engineer must bridge these worlds, not pick one or the other.��Probabilistic Cost Framing: Instead of "this will cost £1.1M", say "there is a 50% probability it will cost less than £1.1M, and a 90% probability it will cost less than £1.65M". This is more honest AND more useful for decision-makers.��Rolling Wave: You don't need the same level of precision for work you're doing in six months as for work starting next sprint. Estimate near-term in detail, longer-horizon at coarser fidelity. Revisit regularly. Accuracy naturally improves as delivery progresses.��Agile EVM: The apparent incompatibility with EVM is a myth. Sprint velocity data gives you the earned value signal. PMI, DAD, and others have published hybrid frameworks. The financial governance visibility programme offices need is achievable without waterfall constraints.��Key message: defending false precision is professionally indefensible. Embrace uncertainty, quantify it rigorously, and communicate it clearly.

12 of 15

06 & 07 | DIGITAL TWINS · REAL-TIME FORECASTING · SUSTAINABILITY

Digital Twin Programme Control

Live Telemetry

Sprint velocity, defect rates, build metrics, actual cost — all flowing continuously

Dynamic Forecasting

Final cost re-forecast continuously — not at monthly reporting cycles

Anomaly Detection

Statistical control limits flag overrun risk before it becomes irreversible

Three Enablers

Data infrastructure · Modelling capability · Governance accountability

The Green Cost Engineer

Carbon as a Cost Dimension

Carbon intensity (gCO₂e/kWh), SCI metrics, PUE — translating technical measures into financial terms for procurement and finance teams.

Regulatory Cost Risk

EU Taxonomy · SEC climate rules · UK SECR · ISSB standards. Programmes ignoring these face material scope and cost growth as obligations crystallise.

The Green Premium Business Case

Renewable-powered regions, energy-efficient code, managed services — quantify carbon savings, regulatory risk mitigation and long-term TCO advantages.

SPEAKER NOTES — Slide 8: Digital Twins & Sustainability��Digital Twins for Programme Control:�The vision: a continuously updated virtual model of your programme, fed by live project telemetry. Not a monthly status report — a real-time financial instrument.�• Programme directors see cost impact of a scope change within hours of it being raised�• Anomaly detection flags when burn rates deviate beyond statistical control limits — early intervention before overruns become irreversible�• Portfolio-level views give CFOs and investment boards real-time visibility across concurrent projects��Three things are required: data infrastructure (reliable telemetry from ALM tools), modelling capability (ML skills to build and maintain dynamic models), and governance (clear accountability for forecast accuracy).��Sustainability:�This is arriving with more urgency than most practitioners have recognised. Software systems are not carbon-neutral.�• Carbon pricing mechanisms are proliferating — these create real financial risk that must be modelled�• The regulatory landscape (EU Taxonomy, SECR, ISSB) will make sustainability cost accounting mandatory for most large organisations within this decade�• The 'green premium' — the extra cost of sustainable architectural choices — needs a rigorous TCO business case, not just a narrative��The cost engineer is uniquely positioned to make this case with financial rigour.

13 of 15

08 | BOLD PREDICTIONS FOR 2030

AI-generated first estimates will be standard practice on major software programmes — humans provide calibration and governance

Function Point counting will be substantially automated through codebase analysis and requirements parsing

Carbon cost will be a mandatory line item in all public sector technology programme business cases

Real-time digital twin forecasting will replace the periodic EV report as the primary cost control mechanism

Quantum computing programmes will require entirely new cost modelling frameworks — a new and lucrative specialisation

The 'Estimation Engineer' role will formally separate from 'Project Controls' in organisational structures

Predictive analytics will routinely identify high-cost-risk programmes before commitment

Cost engineering will be recognised as a C-suite competency — 'Chief Estimation Officer' roles will emerge

SPEAKER NOTES — Slide 9: Bold Predictions for 2030��These are evidence-based projections, not speculation. They are intended to provoke debate and motivate proactive response.��Emphasise that these are not guaranteed outcomes — they will be shaped by the choices practitioners, organisations and professional bodies make in the next few years.��Key predictions to highlight:�• AI-generated estimates as standard practice: the question is not IF but WHEN and HOW governance will be structured around it�• Mandatory carbon cost line items: regulatory direction is clear; the discipline must be ready to deliver this capability�• Digital twin forecasting replacing periodic EV reports: this is already happening in leading organisations�• Chief Estimation Officer: reflects the growing strategic importance of cost credibility at executive level��Ask the audience: which of these is most surprising? Which do they think will arrive fastest? This often generates rich discussion.

14 of 15

08 | FUTURE SKILLS & CALL TO ACTION

The T-Shaped Cost Engineer 2030

Data Science Fluency

Python, ML model evaluation, statistical interpretation — at practitioner not research level

Cloud & FinOps Literacy

Cloud cost structures, consumption-based pricing, FinOps framework

Domain Expertise

Deep knowledge in target domain: enterprise, embedded, AI/ML, quantum

Agile Method Fluency

Participant capability in agile teams — not observer of them

ESG & Sustainability

Identify, quantify and communicate environmental cost dimensions

Communication Skills

Translate probabilistic estimates for boards, audit committees, programme offices

Call to Action

Practitioners

Invest in data science literacy — Python and ML are baseline skills
Build your personal project database systematically
Engage with agile teams as a participant, not a critic
Stay current with ESG and sustainability frameworks

Organisations

Invest in a centralised, governed historical cost repository
Pilot a digital twin cost model on one active programme
Establish AI estimation ethics policy BEFORE deployment at scale

Professional Bodies

Accelerate BoK updates: AI, cloud, agile, sustainability
Develop joint certification pathways with FinOps and data science bodies

SPEAKER NOTES — Slide 10: Future Skills & Call to Action��The T-shaped model is the right mental framework for 2030 cost engineering. Deep in at least one estimation methodology; broad across the full estimation lifecycle. The era of the single-tool specialist is ending.��Six skill areas are now non-negotiable:�1. Data science fluency — not PhD-level, but practitioner-level Python and ML�2. Cloud/FinOps — consumption-based pricing is the new normal; you must speak FinOps�3. Domain expertise — generic estimators are increasingly commoditised�4. Agile fluency — you need to work productively inside agile teams�5. ESG — regulatory obligation within this decade�6. Communication — probabilistic estimates mean nothing if you can't explain them to a programme board��The Call to Action is different for each audience segment:�• Individual practitioners: the investment in skills must start now, not when the job description demands it�• Organisations: governance frameworks and data infrastructure before AI tool adoption, not after�• Professional bodies: the curriculum review cycle must accelerate dramatically — ICEAA, ISMA, IFPUG all have work to do��Close with the final quote from the paper: "The best estimate is not the most precise one. It is the most credible one — built on sound data, rigorous method, honest uncertainty, and the expertise to know the difference."

15 of 15

The Future of Cost Engineering

is Bright —

for those who choose

to shape it.

"The best estimate is not the most precise one. It is the most credible one — built on sound data, rigorous method, honest uncertainty, and the expertise to know the difference."

Sirius Analysis Ltd | Cost Engineering | 2025

Connect on LinkedIn | sirius-analysis.co.uk

SPEAKER NOTES — Slide 11: Closing��Bring the audience back to the central message: cost engineering is not a discipline in decline. It is a discipline in transformation.��The skills, methods and tools that served the profession well through the era of waterfall delivery must now evolve. That evolution is already under way — in research laboratories, in forward-thinking organisations, and in the careers of practitioners who have recognised that standing still is not an option.��The practitioners who will thrive are those who:�• Embrace intellectual curiosity — willing to learn new methods without abandoning disciplined thinking�• Can communicate uncertainty without being accused of evasion�• Can harness the power of AI without losing the critical judgement that remains irreplaceable��Close with an invitation for discussion, questions, and continued conversation via LinkedIn.��Thank the audience for their time and engagement.